From 698a70e6a84cf7c0db667e9d9e021b4c34c85a3e Mon Sep 17 00:00:00 2001 From: m-holger Date: Wed, 24 May 2023 16:28:17 +0100 Subject: Code tidy - reflow comments and strings --- libqpdf/JSON.cc | 68 ++--- libqpdf/Pl_Buffer.cc | 3 +- libqpdf/QPDF.cc | 490 ++++++++++++++-------------------- libqpdf/QPDFAcroFormDocumentHelper.cc | 329 +++++++++-------------- libqpdf/QPDFFormFieldObjectHelper.cc | 63 ++--- libqpdf/QPDFJob.cc | 441 +++++++++++++----------------- libqpdf/QPDFObjectHandle.cc | 88 +++--- libqpdf/QPDFPageObjectHelper.cc | 145 ++++------ libqpdf/QPDFParser.cc | 86 +++--- libqpdf/QPDFTokenizer.cc | 123 ++++----- libqpdf/QPDF_Stream.cc | 56 ++-- libqpdf/QPDF_encryption.cc | 135 ++++------ libqpdf/QPDF_json.cc | 23 +- libqpdf/QPDF_linearization.cc | 437 +++++++++++++----------------- libqpdf/QPDF_optimization.cc | 94 +++---- libqpdf/QPDF_pages.cc | 122 ++++----- 16 files changed, 1118 insertions(+), 1585 deletions(-) (limited to 'libqpdf') diff --git a/libqpdf/JSON.cc b/libqpdf/JSON.cc index e144c7a4..a11d06f2 100644 --- a/libqpdf/JSON.cc +++ b/libqpdf/JSON.cc @@ -218,13 +218,12 @@ JSON::encode_string(std::string const& str) while (iter != end) { auto c = static_cast(*iter); if ((c > 34 && c != '\\') || c == ' ' || c == 33) { - // Optimistically check that no char in str requires escaping. - // Hopefully we can just return the input str. + // Optimistically check that no char in str requires escaping. Hopefully we can just + // return the input str. ++iter; } else { - // We found a char that requires escaping. Initialize result to the - // chars scanned so far, append/replace the rest of str one char at - // a time, and return the result. + // We found a char that requires escaping. Initialize result to the chars scanned so + // far, append/replace the rest of str one char at a time, and return the result. std::string result{begin, iter}; for (; iter != end; ++iter) { @@ -532,12 +531,10 @@ JSON::checkSchemaInternal( } else if (sch_arr) { auto n_elements = sch_arr->elements.size(); if (n_elements == 1) { - // A single-element array in the schema allows a single - // element in the object or a variable-length array, each - // of whose items must conform to the single element of - // the schema array. This doesn't apply to arrays of - // arrays -- we fall back to the behavior of allowing a - // single item only when the object is not an array. + // A single-element array in the schema allows a single element in the object or a + // variable-length array, each of whose items must conform to the single element of the + // schema array. This doesn't apply to arrays of arrays -- we fall back to the behavior + // of allowing a single item only when the object is not an array. if (this_arr) { int i = 0; for (auto const& element: this_arr->elements) { @@ -560,10 +557,9 @@ JSON::checkSchemaInternal( err_prefix + " is supposed to be an array of length " + std::to_string(n_elements)); return false; } else { - // A multi-element array in the schema must correspond to - // an element of the same length in the object. Each - // element in the object is validated against the - // corresponding element in the schema. + // A multi-element array in the schema must correspond to an element of the same length + // in the object. Each element in the object is validated against the corresponding + // element in the schema. size_t i = 0; for (auto const& element: this_arr->elements) { checkSchemaInternal( @@ -701,8 +697,7 @@ JSONParser::handle_u_code( QTC::TC("libtests", "JSON 16 high high"); throw std::runtime_error( "JSON: offset " + std::to_string(new_high_offset) + - ": UTF-16 high surrogate found after previous high surrogate" - " at offset " + + ": UTF-16 high surrogate found after previous high surrogate at offset " + std::to_string(high_offset)); } high_offset = new_high_offset; @@ -713,8 +708,7 @@ JSONParser::handle_u_code( QTC::TC("libtests", "JSON 16 low not after high"); throw std::runtime_error( "JSON: offset " + std::to_string(offset) + - ": UTF-16 low surrogate found not immediately after high" - " surrogate"); + ": UTF-16 low surrogate found not immediately after high surrogate"); } high_offset = 0; codepoint = 0x10000U + ((high_surrogate & 0x3FFU) << 10U) + (codepoint & 0x3FF); @@ -797,8 +791,8 @@ JSONParser::append() ++offset; } -// Append current character to token, advance to next input character and -// transition to 'next' lexer state. +// Append current character to token, advance to next input character and transition to 'next' lexer +// state. inline void JSONParser::append(lex_state_e next) { @@ -808,8 +802,7 @@ JSONParser::append(lex_state_e next) ++offset; } -// Advance to next input character without appending the current character to -// token. +// Advance to next input character without appending the current character to token. inline void JSONParser::ignore() { @@ -817,8 +810,8 @@ JSONParser::ignore() ++offset; } -// Advance to next input character without appending the current character to -// token and transition to 'next' lexer state. +// Advance to next input character without appending the current character to token and transition +// to 'next' lexer state. inline void JSONParser::ignore(lex_state_e next) { @@ -848,9 +841,8 @@ JSONParser::getToken() if ((*p < 32 && *p >= 0)) { if (*p == '\t' || *p == '\n' || *p == '\r') { - // Legal white space not permitted in strings. This will always - // end the current token (unless we are still before the start - // of the token). + // Legal white space not permitted in strings. This will always end the current + // token (unless we are still before the start of the token). if (lex_state == ls_top) { ignore(); } else { @@ -1044,8 +1036,7 @@ JSONParser::getToken() QTC::TC("libtests", "JSON 16 dangling high"); throw std::runtime_error( "JSON: offset " + std::to_string(high_offset) + - ": UTF-16 high surrogate not followed by low " - "surrogate"); + ": UTF-16 high surrogate not followed by low surrogate"); } ignore(); return; @@ -1062,8 +1053,7 @@ JSONParser::getToken() case '\\': case '\"': case '/': - // \/ is allowed in json input, but so is /, so we - // don't map / to \/ in output. + // \/ is allowed in json input, but so is /, so we don't map / to \/ in output. token += *p; break; case 'b': @@ -1113,8 +1103,8 @@ JSONParser::getToken() } } - // We only get here if on end of input or if the last character was a - // control character or other delimiter. + // We only get here if on end of input or if the last character was a control character or other + // delimiter. if (!token.empty()) { switch (lex_state) { @@ -1189,8 +1179,7 @@ JSONParser::handleToken() } else if (parser_state == ps_array_after_item) { parser_state = ps_array_after_comma; } else { - throw std::logic_error("JSONParser::handleToken: unexpected parser" - " state for comma"); + throw std::logic_error("JSONParser::handleToken: unexpected parser state for comma"); } return; @@ -1323,10 +1312,9 @@ JSONParser::handleToken() if (item.isDictionary() || item.isArray()) { stack.push_back({parser_state, item}); - // Calling container start method is postponed until after - // adding the containers to their parent containers, if any. - // This makes it much easier to keep track of the current - // nesting level. + // Calling container start method is postponed until after adding the containers to their + // parent containers, if any. This makes it much easier to keep track of the current nesting + // level. if (item.isDictionary()) { if (reactor) { reactor->dictionaryStart(); diff --git a/libqpdf/Pl_Buffer.cc b/libqpdf/Pl_Buffer.cc index 76c8a5d5..c3184104 100644 --- a/libqpdf/Pl_Buffer.cc +++ b/libqpdf/Pl_Buffer.cc @@ -13,8 +13,7 @@ Pl_Buffer::Pl_Buffer(char const* identifier, Pipeline* next) : Pl_Buffer::~Pl_Buffer() { - // Must be explicit and not inline -- see QPDF_DLL_CLASS in - // README-maintainer + // Must be explicit and not inline -- see QPDF_DLL_CLASS in README-maintainer } void diff --git a/libqpdf/QPDF.cc b/libqpdf/QPDF.cc index 806a8cb2..64ff4715 100644 --- a/libqpdf/QPDF.cc +++ b/libqpdf/QPDF.cc @@ -32,8 +32,8 @@ #include #include -// This must be a fixed value. This API returns a const reference to -// it, and the C API relies on its being static as well. +// This must be a fixed value. This API returns a const reference to it, and the C API relies on its +// being static as well. std::string const QPDF::qpdf_version(QPDF_VERSION); static char const* EMPTY_PDF = ( @@ -212,33 +212,26 @@ QPDF::QPDF() : m(new Members()) { m->tokenizer.allowEOF(); - // Generate a unique ID. It just has to be unique among all QPDF - // objects allocated throughout the lifetime of this running - // application. + // Generate a unique ID. It just has to be unique among all QPDF objects allocated throughout + // the lifetime of this running application. static std::atomic unique_id{0}; m->unique_id = unique_id.fetch_add(1ULL); } QPDF::~QPDF() { - // If two objects are mutually referential (through each object - // having an array or dictionary that contains an indirect - // reference to the other), the circular references in the - // std::shared_ptr objects will prevent the objects from being - // deleted. Walk through all objects in the object cache, which is - // those objects that we read from the file, and break all - // resolved indirect references by replacing them with an internal - // object type representing that they have been destroyed. Note - // that we can't break references like this at any time when the - // QPDF object is active. The call to reset also causes all direct - // QPDFObjectHandle objects that are reachable from this object to - // release their association with this QPDF. Direct objects are - // not destroyed since they can be moved to other QPDF objects - // safely. - - // At this point, obviously no one is still using the QPDF object, - // but we'll explicitly clear the xref table anyway just to - // prevent any possibility of resolve() succeeding. + // If two objects are mutually referential (through each object having an array or dictionary + // that contains an indirect reference to the other), the circular references in the + // std::shared_ptr objects will prevent the objects from being deleted. Walk through all objects + // in the object cache, which is those objects that we read from the file, and break all + // resolved indirect references by replacing them with an internal object type representing that + // they have been destroyed. Note that we can't break references like this at any time when the + // QPDF object is active. The call to reset also causes all direct QPDFObjectHandle objects that + // are reachable from this object to release their association with this QPDF. Direct objects + // are not destroyed since they can be moved to other QPDF objects safely. + + // At this point, obviously no one is still using the QPDF object, but we'll explicitly clear + // the xref table anyway just to prevent any possibility of resolve() succeeding. m->xref_table.clear(); for (auto const& iter: m->obj_cache) { iter.second.object->disconnect(); @@ -406,18 +399,15 @@ QPDF::findHeader() } p += 5; std::string version; - // Note: The string returned by line.c_str() is always - // null-terminated. The code below never overruns the buffer - // because a null character always short-circuits further - // advancement. + // Note: The string returned by line.c_str() is always null-terminated. The code below never + // overruns the buffer because a null character always short-circuits further advancement. bool valid = validatePDFVersion(p, version); if (valid) { m->pdf_version = version; if (global_offset != 0) { - // Empirical evidence strongly suggests that when there is - // leading material prior to the PDF header, all explicit - // offsets in the file are such that 0 points to the - // beginning of the header. + // Empirical evidence strongly suggests that when there is leading material prior to the + // PDF header, all explicit offsets in the file are such that 0 points to the beginning + // of the header. QTC::TC("qpdf", "QPDF global offset"); m->file = std::shared_ptr(new OffsetInputSource(m->file, global_offset)); } @@ -448,14 +438,12 @@ QPDF::parse(char const* password) if (!m->file->findFirst("%PDF-", 0, 1024, hf)) { QTC::TC("qpdf", "QPDF not a pdf file"); warn(damagedPDF("", 0, "can't find PDF header")); - // QPDFWriter writes files that usually require at least - // version 1.2 for /FlateDecode + // QPDFWriter writes files that usually require at least version 1.2 for /FlateDecode m->pdf_version = "1.2"; } - // PDF spec says %%EOF must be found within the last 1024 bytes of - // the file. We add an extra 30 characters to leave room for the - // startxref stuff. + // PDF spec says %%EOF must be found within the last 1024 bytes of/ the file. We add an extra + // 30 characters to leave room for the startxref stuff. m->file->seek(0, SEEK_END); qpdf_offset_t end_offset = m->file->tell(); qpdf_offset_t start_offset = (end_offset > 1054 ? end_offset - 1054 : 0); @@ -494,8 +482,8 @@ void QPDF::inParse(bool v) { if (m->in_parse == v) { - // This happens if QPDFParser::parse tries to - // resolve an indirect object while it is parsing. + // This happens if QPDFParser::parse tries to resolve an indirect object while it is + // parsing. throw std::logic_error("QPDF: re-entrant parsing detected. This is a qpdf bug." " Please report at https://github.com/qpdf/qpdf/issues."); } @@ -518,7 +506,7 @@ QPDF::warn( qpdf_offset_t offset, std::string const& message) { - warn(QPDFExc(error_code, this->getFilename(), object, offset, message)); + warn(QPDFExc(error_code, getFilename(), object, offset, message)); } void @@ -534,9 +522,8 @@ void QPDF::reconstruct_xref(QPDFExc& e) { if (m->reconstructed_xref) { - // Avoid xref reconstruction infinite loops. This is getting - // very hard to reproduce because qpdf is throwing many fewer - // exceptions while parsing. Most situations are warnings now. + // Avoid xref reconstruction infinite loops. This is getting very hard to reproduce because + // qpdf is throwing many fewer exceptions while parsing. Most situations are warnings now. throw e; } @@ -572,8 +559,7 @@ QPDF::reconstruct_xref(QPDFExc& e) QPDFTokenizer::Token t1 = readToken(m->file, MAX_LEN); qpdf_offset_t token_start = m->file->tell() - toO(t1.getValue().length()); if (token_start >= next_line_start) { - // don't process yet -- wait until we get to the line - // containing this token + // don't process yet -- wait until we get to the line containing this token } else if (t1.isInteger()) { QPDFTokenizer::Token t2 = readToken(m->file, MAX_LEN); if ((t2.isInteger()) && (readToken(m->file, MAX_LEN).isWord("obj"))) { @@ -594,22 +580,18 @@ QPDF::reconstruct_xref(QPDFExc& e) } if (!m->trailer.isInitialized()) { - // We could check the last encountered object to see if it was - // an xref stream. If so, we could try to get the trailer - // from there. This may make it possible to recover files - // with bad startxref pointers even when they have object - // streams. + // We could check the last encountered object to see if it was an xref stream. If so, we + // could try to get the trailer from there. This may make it possible to recover files with + // bad startxref pointers even when they have object streams. throw damagedPDF("", 0, "unable to find trailer dictionary while recovering damaged file"); } - // We could iterate through the objects looking for streams and - // try to find objects inside of them, but it's probably not worth - // the trouble. Acrobat can't recover files with any errors in an - // xref stream, and this would be a real long shot anyway. If we - // wanted to do anything that involved looking at stream contents, - // we'd also have to call initializeEncryption() here. It's safe - // to call it more than once. + // We could iterate through the objects looking for streams and try to find objects inside of + // them, but it's probably not worth the trouble. Acrobat can't recover files with any errors + // in an xref stream, and this would be a real long shot anyway. If we wanted to do anything + // that involved looking at stream contents, we'd also have to call initializeEncryption() here. + // It's safe to call it more than once. } void @@ -622,12 +604,10 @@ QPDF::read_xref(qpdf_offset_t xref_offset) char buf[7]; memset(buf, 0, sizeof(buf)); m->file->seek(xref_offset, SEEK_SET); - // Some files miss the mark a little with startxref. We could - // do a better job of searching in the neighborhood for - // something that looks like either an xref table or stream, - // but the simple heuristic of skipping whitespace can help - // with the xref table case and is harmless with the stream - // case. + // Some files miss the mark a little with startxref. We could do a better job of searching + // in the neighborhood for something that looks like either an xref table or stream, but the + // simple heuristic of skipping whitespace can help with the xref table case and is harmless + // with the stream case. bool done = false; bool skipped_space = false; while (!done) { @@ -646,9 +626,8 @@ QPDF::read_xref(qpdf_offset_t xref_offset) } m->file->read(buf, sizeof(buf) - 1); - // The PDF spec says xref must be followed by a line - // terminator, but files exist in the wild where it is - // terminated by arbitrary whitespace. + // The PDF spec says xref must be followed by a line terminator, but files exist in the wild + // where it is terminated by arbitrary whitespace. if ((strncmp(buf, "xref", 4) == 0) && QUtil::is_space(buf[4])) { if (skipped_space) { QTC::TC("qpdf", "QPDF xref skipped space"); @@ -662,8 +641,7 @@ QPDF::read_xref(qpdf_offset_t xref_offset) : (buf[4] == ' ') ? 2 : 9999)); int skip = 4; - // buf is null-terminated, and QUtil::is_space('\0') is - // false, so this won't overrun. + // buf is null-terminated, and QUtil::is_space('\0') is false, so this won't overrun. while (QUtil::is_space(buf[skip])) { ++skip; } @@ -697,16 +675,16 @@ QPDF::read_xref(qpdf_offset_t xref_offset) ") is not one plus the highest object number (" + std::to_string(max_obj) + ")"))); } - // We no longer need the deleted_objects table, so go ahead and - // clear it out to make sure we never depend on its being set. + // We no longer need the deleted_objects table, so go ahead and clear it out to make sure we + // never depend on its being set. m->deleted_objects.clear(); } bool QPDF::parse_xrefFirst(std::string const& line, int& obj, int& num, int& bytes) { - // is_space and is_digit both return false on '\0', so this will - // not overrun the null-terminated buffer. + // is_space and is_digit both return false on '\0', so this will not overrun the null-terminated + // buffer. char const* p = line.c_str(); char const* start = line.c_str(); @@ -753,8 +731,8 @@ QPDF::parse_xrefFirst(std::string const& line, int& obj, int& num, int& bytes) bool QPDF::parse_xrefEntry(std::string const& line, qpdf_offset_t& f1, int& f2, char& type) { - // is_space and is_digit both return false on '\0', so this will - // not overrun the null-terminated buffer. + // is_space and is_digit both return false on '\0', so this will not overrun the null-terminated + // buffer. char const* p = line.c_str(); // Skip zero or more spaces. There aren't supposed to be any. @@ -862,8 +840,7 @@ QPDF::read_xrefTable(qpdf_offset_t xref_offset) "xref table", "invalid xref entry (obj=" + std::to_string(i) + ")"); } if (type == 'f') { - // Save deleted items until after we've checked the - // XRefStm, if any. + // Save deleted items until after we've checked the XRefStm, if any. deleted_items.push_back(QPDFObjGen(toI(i), f2)); } else { insertXrefEntry(toI(i), 1, f1, f2); @@ -902,9 +879,8 @@ QPDF::read_xrefTable(qpdf_offset_t xref_offset) QTC::TC("qpdf", "QPDF ignoring XRefStm in trailer"); } else { if (cur_trailer.getKey("/XRefStm").isInteger()) { - // Read the xref stream but disregard any return value - // -- we'll use our trailer's /Prev key instead of the - // xref stream's. + // Read the xref stream but disregard any return value -- we'll use our trailer's + // /Prev key instead of the xref stream's. (void)read_xrefStream(cur_trailer.getKey("/XRefStm").getIntValue()); } else { throw damagedPDF("xref stream", xref_offset, "invalid /XRefStm"); @@ -1035,8 +1011,8 @@ QPDF::processXRefStream(qpdf_offset_t xref_offset, QPDFObjectHandle& xref_obj) num_entries += toS(indx.at(i)); } - // entry_size and num_entries have both been validated to ensure - // that this multiplication does not cause an overflow. + // entry_size and num_entries have both been validated to ensure that this multiplication does + // not cause an overflow. size_t expected_size = entry_size * num_entries; std::shared_ptr bp = xref_obj.getStreamData(qpdf_dl_specialized); @@ -1060,9 +1036,8 @@ QPDF::processXRefStream(qpdf_offset_t xref_offset, QPDFObjectHandle& xref_obj) bool saw_first_compressed_object = false; - // Actual size vs. expected size check above ensures that we will - // not overflow any buffers here. We know that entry_size * - // num_entries is equal to the size of the buffer. + // Actual size vs. expected size check above ensures that we will not overflow any buffers here. + // We know that entry_size * num_entries is equal to the size of the buffer. unsigned char const* data = bp->getBuffer(); for (size_t i = 0; i < num_entries; ++i) { // Read this entry @@ -1081,17 +1056,15 @@ QPDF::processXRefStream(qpdf_offset_t xref_offset, QPDFObjectHandle& xref_obj) } } - // Get the object and generation number. The object number is - // based on /Index. The generation number is 0 unless this is - // an uncompressed object record, in which case the generation - // number appears as the third field. + // Get the object and generation number. The object number is based on /Index. The + // generation number is 0 unless this is an uncompressed object record, in which case the + // generation number appears as the third field. int obj = toI(indx.at(cur_chunk)); if ((obj < 0) || ((std::numeric_limits::max() - obj) < chunk_count)) { std::ostringstream msg; msg.imbue(std::locale::classic()); msg << "adding " << chunk_count << " to " << obj - << " while computing index in xref stream would cause" - << " an integer overflow"; + << " while computing index in xref stream would cause an integer overflow"; throw std::range_error(msg.str()); } obj += chunk_count; @@ -1113,10 +1086,8 @@ QPDF::processXRefStream(qpdf_offset_t xref_offset, QPDFObjectHandle& xref_obj) m->first_xref_item_offset = xref_offset; } if (fields[0] == 0) { - // Ignore fields[2], which we don't care about in this - // case. This works around the issue of some PDF files - // that put invalid values, like -1, here for deleted - // objects. + // Ignore fields[2], which we don't care about in this case. This works around the issue + // of some PDF files that put invalid values, like -1, here for deleted objects. fields[2] = 0; } insertXrefEntry(obj, toI(fields[0]), fields[1], toI(fields[2])); @@ -1143,17 +1114,14 @@ QPDF::processXRefStream(qpdf_offset_t xref_offset, QPDFObjectHandle& xref_obj) void QPDF::insertXrefEntry(int obj, int f0, qpdf_offset_t f1, int f2, bool overwrite) { - // Populate the xref table in such a way that the first reference - // to an object that we see, which is the one in the latest xref - // table in which it appears, is the one that gets stored. This - // works because we are reading more recent appends before older - // ones. Exception: if overwrite is true, then replace any - // existing object. This is used in xref recovery mode, which - // reads the file from beginning to end. - - // If there is already an entry for this object and generation in - // the table, it means that a later xref table has registered this - // object. Disregard this one. + // Populate the xref table in such a way that the first reference to an object that we see, + // which is the one in the latest xref table in which it appears, is the one that gets stored. + // This works because we are reading more recent appends before older ones. Exception: if + // overwrite is true, then replace any existing object. This is used in xref recovery mode, + // which reads the file from beginning to end. + + // If there is already an entry for this object and generation in the table, it means that a + // later xref table has registered this object. Disregard this one. { // private scope int gen = (f0 == 2 ? 0 : f2); QPDFObjGen og(obj, gen); @@ -1220,8 +1188,8 @@ QPDF::showXRefTable() } } -// Resolve all objects in the xref table. If this triggers a xref table -// reconstruction abort and return false. Otherwise return true. +// Resolve all objects in the xref table. If this triggers a xref table reconstruction abort and +// return false. Otherwise return true. bool QPDF::resolveXRefTable() { @@ -1237,8 +1205,8 @@ QPDF::resolveXRefTable() return true; } -// Ensure all objects in the pdf file, including those in indirect -// references, appear in the object cache. +// Ensure all objects in the pdf file, including those in indirect references, appear in the object +// cache. void QPDF::fixDanglingReferences(bool force) { @@ -1255,10 +1223,9 @@ QPDF::fixDanglingReferences(bool force) size_t QPDF::getObjectCount() { - // This method returns the next available indirect object number. - // makeIndirectObject uses it for this purpose. After - // fixDanglingReferences is called, all objects in the xref table - // will also be in obj_cache. + // This method returns the next available indirect object number. makeIndirectObject uses it for + // this purpose. After fixDanglingReferences is called, all objects in the xref table will also + // be in obj_cache. fixDanglingReferences(); QPDFObjGen og; if (!m->obj_cache.empty()) { @@ -1270,8 +1237,7 @@ QPDF::getObjectCount() std::vector QPDF::getAllObjects() { - // After fixDanglingReferences is called, all objects are in the - // object cache. + // After fixDanglingReferences is called, all objects are in the object cache. fixDanglingReferences(); std::vector result; for (auto const& iter: m->obj_cache) { @@ -1315,34 +1281,27 @@ QPDF::readObject( auto object = QPDFParser(input, m->last_object_description, m->tokenizer, decrypter, this) .parse(empty, false); if (empty) { - // Nothing in the PDF spec appears to allow empty objects, but - // they have been encountered in actual PDF files and Adobe - // Reader appears to ignore them. + // Nothing in the PDF spec appears to allow empty objects, but they have been encountered in + // actual PDF files and Adobe Reader appears to ignore them. warn(damagedPDF(input, input->getLastOffset(), "empty object treated as null")); } else if (object.isDictionary() && (!in_object_stream)) { // check for stream qpdf_offset_t cur_offset = input->tell(); if (readToken(input).isWord("stream")) { - // The PDF specification states that the word "stream" - // should be followed by either a carriage return and - // a newline or by a newline alone. It specifically - // disallowed following it by a carriage return alone - // since, in that case, there would be no way to tell - // whether the NL in a CR NL sequence was part of the - // stream data. However, some readers, including - // Adobe reader, accept a carriage return by itself - // when followed by a non-newline character, so that's - // what we do here. We have also seen files that have - // extraneous whitespace between the stream keyword and - // the newline. + // The PDF specification states that the word "stream" should be followed by either a + // carriage return and a newline or by a newline alone. It specifically disallowed + // following it by a carriage return alone since, in that case, there would be no way to + // tell whether the NL in a CR NL sequence was part of the stream data. However, some + // readers, including Adobe reader, accept a carriage return by itself when followed by + // a non-newline character, so that's what we do here. We have also seen files that have + // extraneous whitespace between the stream keyword and the newline. bool done = false; while (!done) { done = true; char ch; if (input->read(&ch, 1) == 0) { - // A premature EOF here will result in some - // other problem that will get reported at - // another time. + // A premature EOF here will result in some other problem that will get reported + // at another time. } else if (ch == '\n') { // ready to read stream data QTC::TC("qpdf", "QPDF stream with NL only"); @@ -1353,10 +1312,8 @@ QPDF::readObject( // Ready to read stream data QTC::TC("qpdf", "QPDF stream with CRNL"); } else { - // Treat the \r by itself as the - // whitespace after endstream and - // start reading stream data in spite - // of not having seen a newline. + // Treat the \r by itself as the whitespace after endstream and start + // reading stream data in spite of not having seen a newline. QTC::TC("qpdf", "QPDF stream with CR only"); input->unreadCh(ch); warn(damagedPDF( @@ -1381,9 +1338,8 @@ QPDF::readObject( } } - // Must get offset before accessing any additional - // objects since resolving a previously unresolved - // indirect object will change file position. + // Must get offset before accessing any additional objects since resolving a previously + // unresolved indirect object will change file position. qpdf_offset_t stream_offset = input->tell(); size_t length = 0; @@ -1427,8 +1383,7 @@ QPDF::readObject( } } - // Override last_offset so that it points to the beginning of the - // object we just read + // Override last_offset so that it points to the beginning of the object we just read input->setLastOffset(offset); return object; } @@ -1449,8 +1404,7 @@ size_t QPDF::recoverStreamLength( std::shared_ptr input, QPDFObjGen const& og, qpdf_offset_t stream_offset) { - // Try to reconstruct stream length by looking for - // endstream or endobj + // Try to reconstruct stream length by looking for endstream or endobj warn(damagedPDF(input, stream_offset, "attempting to recover stream length")); PatternFinder ef(*this, &QPDF::findEndstream); @@ -1481,9 +1435,8 @@ QPDF::recoverStreamLength( } } if (this_obj_offset && (this_og == og)) { - // Well, we found endstream\nendobj within the space - // allowed for this object, so we're probably in good - // shape. + // Well, we found endstream\nendobj within the space allowed for this object, so we're + // probably in good shape. } else { QTC::TC("qpdf", "QPDF found wrong endstream in recovery"); } @@ -1518,14 +1471,12 @@ QPDF::readObjectAtOffset( { bool check_og = true; if (exp_og.getObj() == 0) { - // This method uses an expect object ID of 0 to indicate that - // we don't know or don't care what the actual object ID is at - // this offset. This is true when we read the xref stream and - // linearization hint streams. In this case, we don't verify - // the expect object ID/generation against what was read from - // the file. There is also no reason to attempt xref recovery - // if we get a failure in this case since the read attempt was - // not triggered by an xref lookup. + // This method uses an expect object ID of 0 to indicate that we don't know or don't care + // what the actual object ID is at this offset. This is true when we read the xref stream + // and linearization hint streams. In this case, we don't verify the expect object + // ID/generation against what was read from the file. There is also no reason to attempt + // xref recovery if we get a failure in this case since the read attempt was not triggered + // by an xref lookup. check_og = false; try_recovery = false; } @@ -1535,11 +1486,9 @@ QPDF::readObjectAtOffset( try_recovery = false; } - // Special case: if offset is 0, just return null. Some PDF - // writers, in particular "Mac OS X 10.7.5 Quartz PDFContext", may - // store deleted objects in the xref table as "0000000000 00000 - // n", which is not correct, but it won't hurt anything for to - // ignore these. + // Special case: if offset is 0, just return null. Some PDF writers, in particular + // "Mac OS X 10.7.5 Quartz PDFContext", may store deleted objects in the xref table as + // "0000000000 00000 n", which is not correct, but it won't hurt anything for to ignore these. if (offset == 0) { QTC::TC("qpdf", "QPDF bogus 0 offset", 0); warn(damagedPDF(0, "object has offset 0")); @@ -1579,8 +1528,7 @@ QPDF::readObjectAtOffset( // Will be retried below throw e; } else { - // We can try reading the object anyway even if the ID - // doesn't match. + // We can try reading the object anyway even if the ID doesn't match. warn(e); } } @@ -1617,16 +1565,13 @@ QPDF::readObjectAtOffset( } if (isUnresolved(og)) { - // Store the object in the cache here so it gets cached - // whether we first know the offset or whether we first know - // the object ID and generation (in which we case we would get - // here through resolve). - - // Determine the end offset of this object before and after - // white space. We use these numbers to validate - // linearization hint tables. Offsets and lengths of objects - // may imply the end of an object to be anywhere between these - // values. + // Store the object in the cache here so it gets cached whether we first know the offset or + // whether we first know the object ID and generation (in which we case we would get here + // through resolve). + + // Determine the end offset of this object before and after white space. We use these + // numbers to validate linearization hint tables. Offsets and lengths of objects may imply + // the end of an object to be anywhere between these values. qpdf_offset_t end_before_space = m->file->tell(); // skip over spaces @@ -1643,41 +1588,31 @@ QPDF::readObjectAtOffset( } qpdf_offset_t end_after_space = m->file->tell(); if (skip_cache_if_in_xref && m->xref_table.count(og)) { - // Ordinarily, an object gets read here when resolved - // through xref table or stream. In the special case of - // the xref stream and linearization hint tables, the - // offset comes from another source. For the specific case - // of xref streams, the xref stream is read and loaded - // into the object cache very early in parsing. - // Ordinarily, when a file is updated by appending, items - // inserted into the xref table in later updates take - // precedence over earlier items. In the special case of - // reusing the object number previously used as the xref - // stream, we have the following order of events: + // Ordinarily, an object gets read here when resolved through xref table or stream. In + // the special case of the xref stream and linearization hint tables, the offset comes + // from another source. For the specific case of xref streams, the xref stream is read + // and loaded into the object cache very early in parsing. Ordinarily, when a file is + // updated by appending, items inserted into the xref table in later updates take + // precedence over earlier items. In the special case of reusing the object number + // previously used as the xref stream, we have the following order of events: // // * reused object gets loaded into the xref table // * old object is read here while reading xref streams // * original xref entry is ignored (since already in xref table) // - // It is the second step that causes a problem. Even - // though the xref table is correct in this case, the old - // object is already in the cache and so effectively - // prevails over the reused object. To work around this - // issue, we have a special case for the xref stream (via - // the skip_cache_if_in_xref): if the object is already in - // the xref stream, don't cache what we read here. + // It is the second step that causes a problem. Even though the xref table is correct in + // this case, the old object is already in the cache and so effectively prevails over + // the reused object. To work around this issue, we have a special case for the xref + // stream (via the skip_cache_if_in_xref): if the object is already in the xref stream, + // don't cache what we read here. // - // It is likely that the same bug may exist for - // linearization hint tables, but the existing code uses - // end_before_space and end_after_space from the cache, so - // fixing that would require more significant rework. The - // chances of a linearization hint stream being reused - // seems smaller because the xref stream is probably the - // highest object in the file and the linearization hint - // stream would be some random place in the middle, so I'm - // leaving that bug unfixed for now. If the bug were to be - // fixed, we could use !check_og in place of - // skip_cache_if_in_xref. + // It is likely that the same bug may exist for linearization hint tables, but the + // existing code uses end_before_space and end_after_space from the cache, so fixing + // that would require more significant rework. The chances of a linearization hint + // stream being reused seems smaller because the xref stream is probably the highest + // object in the file and the linearization hint stream would be some random place in + // the middle, so I'm leaving that bug unfixed for now. If the bug were to be fixed, we + // could use !check_og in place of skip_cache_if_in_xref. QTC::TC("qpdf", "QPDF skipping cache for known unchecked object"); } else { updateCache(og, oh.getObj(), end_before_space, end_after_space); @@ -1695,9 +1630,8 @@ QPDF::resolve(QPDFObjGen og) } if (m->resolving.count(og)) { - // This can happen if an object references itself directly or - // indirectly in some key that has to be resolved during - // object parsing, such as stream length. + // This can happen if an object references itself directly or indirectly in some key that + // has to be resolved during object parsing, such as stream length. QTC::TC("qpdf", "QPDF recursion loop in resolve"); warn(damagedPDF("", "loop detected resolving object " + og.unparse(' '))); updateCache(og, QPDF_Null::create(), -1, -1); @@ -1758,8 +1692,8 @@ QPDF::resolveObjectsInStream(int obj_stream_number) "supposed object stream " + std::to_string(obj_stream_number) + " is not a stream"); } - // For linearization data in the object, use the data from the - // object stream for the objects in the stream. + // For linearization data in the object, use the data from the object stream for the objects in + // the stream. QPDFObjGen stream_og(obj_stream_number, 0); qpdf_offset_t end_before_space = m->obj_cache[stream_og].end_before_space; qpdf_offset_t end_after_space = m->obj_cache[stream_og].end_after_space; @@ -1804,11 +1738,10 @@ QPDF::resolveObjectsInStream(int obj_stream_number) offsets[num] = toI(offset + first); } - // To avoid having to read the object stream multiple times, store - // all objects that would be found here in the cache. Remember - // that some objects stored here might have been overridden by new - // objects appended to the file, so it is necessary to recheck the - // xref table and only cache what would actually be resolved here. + // To avoid having to read the object stream multiple times, store all objects that would be + // found here in the cache. Remember that some objects stored here might have been overridden + // by new objects appended to the file, so it is necessary to recheck the xref table and only + // cache what would actually be resolved here. for (auto const& iter: offsets) { QPDFObjGen og(iter.first, 0); QPDFXRefEntry const& entry = m->xref_table[og]; @@ -1936,8 +1869,7 @@ QPDF::reserveStream(QPDFObjGen const& og) QPDFObjectHandle QPDF::getObject(QPDFObjGen const& og) { - // This method is called by the parser and therefore must not - // resolve any objects. + // This method is called by the parser and therefore must not resolve any objects. if (!isCached(og)) { m->obj_cache[og] = ObjCache(QPDF_Unresolved::create(this, og), -1, -1); } @@ -1991,48 +1923,38 @@ QPDF::copyForeignObject(QPDFObjectHandle foreign) { // Here's an explanation of what's going on here. // - // A QPDFObjectHandle that is an indirect object has an owning - // QPDF. The object ID and generation refers to an object in the - // owning QPDF. When we copy the QPDFObjectHandle from a foreign - // QPDF into the local QPDF, we have to replace all indirect - // object references with references to the corresponding object - // in the local file. + // A QPDFObjectHandle that is an indirect object has an owning QPDF. The object ID and + // generation refers to an object in the owning QPDF. When we copy the QPDFObjectHandle from a + // foreign QPDF into the local QPDF, we have to replace all indirect object references with + // references to the corresponding object in the local file. // - // To do this, we maintain mappings from foreign object IDs to - // local object IDs for each foreign QPDF that we are copying - // from. The mapping is stored in an ObjCopier, which contains a + // To do this, we maintain mappings from foreign object IDs to local object IDs for each foreign + // QPDF that we are copying from. The mapping is stored in an ObjCopier, which contains a // mapping from the foreign ObjGen to the local QPDFObjectHandle. // - // To copy, we do a deep traversal of the foreign object with loop - // detection to discover all indirect objects that are - // encountered, stopping at page boundaries. Whenever we encounter - // an indirect object, we check to see if we have already created - // a local copy of it. If not, we allocate a "reserved" object - // (or, for a stream, just a new stream) and store in the map the + // To copy, we do a deep traversal of the foreign object with loop detection to discover all + // indirect objects that are encountered, stopping at page boundaries. Whenever we encounter an + // indirect object, we check to see if we have already created a local copy of it. If not, we + // allocate a "reserved" object (or, for a stream, just a new stream) and store in the map the // mapping from the foreign object ID to the new object. While we // do this, we keep a list of objects to copy. // - // Once we are done with the traversal, we copy all the objects - // that we need to copy. However, the copies will contain indirect - // object IDs that refer to objects in the foreign file. We need - // to replace them with references to objects in the local file. - // This is what replaceForeignIndirectObjects does. Once we have - // created a copy of the foreign object with all the indirect - // references replaced with new ones in the local context, we can - // replace the local reserved object with the copy. This mechanism - // allows us to copy objects with circular references in any - // order. - - // For streams, rather than copying the objects, we set up the - // stream data to pull from the original stream by using a stream - // data provider. This is done in a manner that doesn't require - // the original QPDF object but may require the original source of - // the stream data with special handling for immediate_copy_from. - // This logic is also in replaceForeignIndirectObjects. - - // Note that we explicitly allow use of copyForeignObject on page - // objects. It is a documented use case to copy pages this way if - // the intention is to not update the pages tree. + // Once we are done with the traversal, we copy all the objects that we need to copy. However, + // the copies will contain indirect object IDs that refer to objects in the foreign file. We + // need to replace them with references to objects in the local file. This is what + // replaceForeignIndirectObjects does. Once we have created a copy of the foreign object with + // all the indirect references replaced with new ones in the local context, we can replace the + // local reserved object with the copy. This mechanism allows us to copy objects with circular + // references in any order. + + // For streams, rather than copying the objects, we set up the stream data to pull from the + // original stream by using a stream data provider. This is done in a manner that doesn't + // require the original QPDF object but may require the original source of the stream data with + // special handling for immediate_copy_from. This logic is also in + // replaceForeignIndirectObjects. + + // Note that we explicitly allow use of copyForeignObject on page objects. It is a documented + // use case to copy pages this way if the intention is to not update the pages tree. if (!foreign.isIndirect()) { QTC::TC("qpdf", "QPDF copyForeign direct"); throw std::logic_error("QPDF::copyForeign called with direct object handle"); @@ -2049,12 +1971,10 @@ QPDF::copyForeignObject(QPDFObjectHandle foreign) " at the beginning of copyForeignObject"); } - // Make sure we have an object in this file for every referenced - // object in the old file. obj_copier.object_map maps foreign - // QPDFObjGen to local objects. For everything new that we have - // to copy, the local object will be a reservation, unless it is a - // stream, in which case the local object will already be a - // stream. + // Make sure we have an object in this file for every referenced object in the old file. + // obj_copier.object_map maps foreign QPDFObjGen to local objects. For everything new that we + // have to copy, the local object will be a reservation, unless it is a stream, in which case + // the local object will already be a stream. reserveObjects(foreign, obj_copier, true); if (!obj_copier.visiting.empty()) { @@ -2140,8 +2060,8 @@ QPDF::replaceForeignIndirectObjects(QPDFObjectHandle foreign, ObjCopier& obj_cop QTC::TC("qpdf", "QPDF replace indirect"); auto mapping = obj_copier.object_map.find(foreign.getObjGen()); if (mapping == obj_copier.object_map.end()) { - // This case would occur if this is a reference to a Page - // or Pages object that we didn't traverse into. + // This case would occur if this is a reference to a Page or Pages object that we didn't + // traverse into. QTC::TC("qpdf", "QPDF replace foreign indirect with null"); result = QPDFObjectHandle::newNull(); } else { @@ -2192,9 +2112,8 @@ QPDF::replaceForeignIndirectObjects(QPDFObjectHandle foreign, ObjCopier& obj_cop void QPDF::copyStreamData(QPDFObjectHandle result, QPDFObjectHandle foreign) { - // This method was originally written for copying foreign streams, - // but it is used by QPDFObjectHandle to copy streams from the - // same QPDF object as well. + // This method was originally written for copying foreign streams, but it is used by + // QPDFObjectHandle to copy streams from the same QPDF object as well. QPDFObjectHandle dict = result.getDict(); QPDFObjectHandle old_dict = foreign.getDict(); @@ -2204,8 +2123,8 @@ QPDF::copyStreamData(QPDFObjectHandle result, QPDFObjectHandle foreign) std::shared_ptr(m->copied_stream_data_provider); } QPDFObjGen local_og(result.getObjGen()); - // Copy information from the foreign stream so we can pipe its - // data later without keeping the original QPDF object around. + // Copy information from the foreign stream so we can pipe its data later without keeping the + // original QPDF object around. QPDF& foreign_stream_qpdf = foreign.getQPDF("unable to retrieve owning qpdf from foreign stream"); @@ -2217,10 +2136,9 @@ QPDF::copyStreamData(QPDFObjectHandle result, QPDFObjectHandle foreign) } std::shared_ptr stream_buffer = stream->getStreamDataBuffer(); if ((foreign_stream_qpdf.m->immediate_copy_from) && (stream_buffer == nullptr)) { - // Pull the stream data into a buffer before attempting - // the copy operation. Do it on the source stream so that - // if the source stream is copied multiple times, we don't - // have to keep duplicating the memory. + // Pull the stream data into a buffer before attempting the copy operation. Do it on the + // source stream so that if the source stream is copied multiple times, we don't have to + // keep duplicating the memory. QTC::TC("qpdf", "QPDF immediate copy stream data"); foreign.replaceStreamData( foreign.getRawStreamData(), @@ -2263,8 +2181,7 @@ QPDF::swapObjects(int objid1, int generation1, int objid2, int generation2) void QPDF::swapObjects(QPDFObjGen const& og1, QPDFObjGen const& og2) { - // Force objects to be read from the input source if needed, then - // swap them in the cache. + // Force objects to be read from the input source if needed, then swap them in the cache. resolve(og1); resolve(og2); m->obj_cache[og1].object->swapWith(m->obj_cache[og2].object); @@ -2338,9 +2255,8 @@ QPDF::getRoot() if (!root.isDictionary()) { throw damagedPDF("", 0, "unable to find /Root dictionary"); } else if ( - // Check_mode is an interim solution to request #810 pending a more - // comprehensive review of the approach to more extensive checks and - // warning levels. + // Check_mode is an interim solution to request #810 pending a more comprehensive review of + // the approach to more extensive checks and warning levels. m->check_mode && !root.getKey("/Type").isNameAndEquals("/Catalog")) { warn(damagedPDF("", 0, "catalog /Type entry missing or invalid")); root.replaceKey("/Type", "/Catalog"_qpdf); @@ -2373,14 +2289,11 @@ QPDF::getObjectStreamData(std::map& omap) std::vector QPDF::getCompressibleObjGens() { - // Return a list of objects that are allowed to be in object - // streams. Walk through the objects by traversing the document - // from the root, including a traversal of the pages tree. This - // makes that objects that are on the same page are more likely to - // be in the same object stream, which is slightly more efficient, - // particularly with linearized files. This is better than - // iterating through the xref table since it avoids preserving - // orphaned items. + // Return a list of objects that are allowed to be in object streams. Walk through the objects + // by traversing the document from the root, including a traversal of the pages tree. This + // makes that objects that are on the same page are more likely to be in the same object stream, + // which is slightly more efficient, particularly with linearized files. This is better than + // iterating through the xref table since it avoids preserving orphaned items. // Exclude encryption dictionary, if any QPDFObjectHandle encryption_dict = m->trailer.getKey("/Encrypt"); @@ -2555,9 +2468,8 @@ QPDF::pipeForeignStreamData( will_retry); } -// Throw a generic exception when we lack context for something -// more specific. New code should not use this. This method exists -// to improve somewhat from calling assert in very old code. +// Throw a generic exception when we lack context for something more specific. New code should not +// use this. This method exists to improve somewhat from calling assert in very old code. void QPDF::stopOnError(std::string const& message) { @@ -2584,33 +2496,31 @@ QPDF::damagedPDF( return damagedPDF(input, m->last_object_description, offset, message); } -// Return an exception of type qpdf_e_damaged_pdf. The filename is taken from -// m->file. +// Return an exception of type qpdf_e_damaged_pdf. The filename is taken from m->file. QPDFExc QPDF::damagedPDF(std::string const& object, qpdf_offset_t offset, std::string const& message) { return QPDFExc(qpdf_e_damaged_pdf, m->file->getName(), object, offset, message); } -// Return an exception of type qpdf_e_damaged_pdf. The filename is taken from -// m->file and the offset from .m->file->getLastOffset(). +// Return an exception of type qpdf_e_damaged_pdf. The filename is taken from m->file and the +// offset from .m->file->getLastOffset(). QPDFExc QPDF::damagedPDF(std::string const& object, std::string const& message) { return damagedPDF(object, m->file->getLastOffset(), message); } -// Return an exception of type qpdf_e_damaged_pdf. The filename is taken from -// m->file and the object from .m->last_object_description. +// Return an exception of type qpdf_e_damaged_pdf. The filename is taken from m->file and the object +// from .m->last_object_description. QPDFExc QPDF::damagedPDF(qpdf_offset_t offset, std::string const& message) { return damagedPDF(m->last_object_description, offset, message); } -// Return an exception of type qpdf_e_damaged_pdf. The filename is taken from -// m->file, the object from m->last_object_description and the offset from -// m->file->getLastOffset(). +// Return an exception of type qpdf_e_damaged_pdf. The filename is taken from m->file, the object +// from m->last_object_description and the offset from m->file->getLastOffset(). QPDFExc QPDF::damagedPDF(std::string const& message) { diff --git a/libqpdf/QPDFAcroFormDocumentHelper.cc b/libqpdf/QPDFAcroFormDocumentHelper.cc index d210d50a..aca4bf1f 100644 --- a/libqpdf/QPDFAcroFormDocumentHelper.cc +++ b/libqpdf/QPDFAcroFormDocumentHelper.cc @@ -15,9 +15,8 @@ QPDFAcroFormDocumentHelper::QPDFAcroFormDocumentHelper(QPDF& qpdf) : QPDFDocumentHelper(qpdf), m(new Members()) { - // We have to analyze up front. Otherwise, when we are adding - // annotations and fields, we are in a temporarily unstable - // configuration where some widget annotations are not reachable. + // We have to analyze up front. Otherwise, when we are adding annotations and fields, we are in + // a temporarily unstable configuration where some widget annotations are not reachable. analyze(); } @@ -77,14 +76,11 @@ QPDFAcroFormDocumentHelper::addAndRenameFormFields(std::vector } if (obj.hasKey("/T")) { - // Find something we can append to the partial name that - // makes the fully qualified name unique. When we find - // something, reuse the same suffix for all fields in this - // group with the same name. We can only change the name - // of fields that have /T, and this field's /T is always - // at the end of the fully qualified name, appending to /T - // has the effect of appending the same thing to the fully - // qualified name. + // Find something we can append to the partial name that makes the fully qualified + // name unique. When we find something, reuse the same suffix for all fields in this + // group with the same name. We can only change the name of fields that have /T, and + // this field's /T is always at the end of the fully qualified name, appending to /T + // has the effect of appending the same thing to the fully qualified name. std::string old_name = QPDFFormFieldObjectHelper(obj).getFullyQualifiedName(); if (renames.count(old_name) == 0) { std::string new_name = old_name; @@ -253,8 +249,7 @@ QPDFAcroFormDocumentHelper::analyze() fields = QPDFObjectHandle::newArray(); } - // Traverse /AcroForm to find annotations and map them - // bidirectionally to fields. + // Traverse /AcroForm to find annotations and map them bidirectionally to fields. QPDFObjGen::set visited; int nfields = fields.getArrayNItems(); @@ -263,12 +258,10 @@ QPDFAcroFormDocumentHelper::analyze() traverseField(fields.getArrayItem(i), null, 0, visited); } - // All Widget annotations should have been encountered by - // traversing /AcroForm, but in case any weren't, find them by - // walking through pages, and treat any widget annotation that is - // not associated with a field as its own field. This just ensures - // that requesting the field for any annotation we find through a - // page's /Annots list will have some associated field. Note that + // All Widget annotations should have been encountered by traversing /AcroForm, but in case any + // weren't, find them by walking through pages, and treat any widget annotation that is not + // associated with a field as its own field. This just ensures that requesting the field for any + // annotation we find through a page's /Annots list will have some associated field. Note that // a file that contains this kind of error will probably not // actually work with most viewers. @@ -278,13 +271,11 @@ QPDFAcroFormDocumentHelper::analyze() QPDFObjGen og(annot.getObjGen()); if (m->annotation_to_field.count(og) == 0) { QTC::TC("qpdf", "QPDFAcroFormDocumentHelper orphaned widget"); - // This is not supposed to happen, but it's easy - // enough for us to handle this case. Treat the - // annotation as its own field. This could allow qpdf - // to sensibly handle a case such as a PDF creator - // adding a self-contained annotation (merged with the - // field dictionary) to the page's /Annots array and - // forgetting to also put it in /AcroForm. + // This is not supposed to happen, but it's easy enough for us to handle this case. + // Treat the annotation as its own field. This could allow qpdf to sensibly handle a + // case such as a PDF creator adding a self-contained annotation (merged with the + // field dictionary) to the page's /Annots array and forgetting to also put it in + // /AcroForm. annot.warnIfPossible("this widget annotation is not" " reachable from /AcroForm in the document catalog"); m->annotation_to_field[og] = QPDFFormFieldObjectHelper(annot); @@ -299,14 +290,14 @@ QPDFAcroFormDocumentHelper::traverseField( QPDFObjectHandle field, QPDFObjectHandle parent, int depth, QPDFObjGen::set& visited) { if (depth > 100) { - // Arbitrarily cut off recursion at a fixed depth to avoid - // specially crafted files that could cause stack overflow. + // Arbitrarily cut off recursion at a fixed depth to avoid specially crafted files that + // could cause stack overflow. return; } if (!field.isIndirect()) { QTC::TC("qpdf", "QPDFAcroFormDocumentHelper direct field"); - field.warnIfPossible("encountered a direct object as a field or annotation while" - " traversing /AcroForm; ignoring field or annotation"); + field.warnIfPossible("encountered a direct object as a field or annotation while " + "traversing /AcroForm; ignoring field or annotation"); return; } if (!field.isDictionary()) { @@ -322,13 +313,11 @@ QPDFAcroFormDocumentHelper::traverseField( return; } - // A dictionary encountered while traversing the /AcroForm field - // may be a form field, an annotation, or the merger of the two. A - // field that has no fields below it is a terminal. If a terminal - // field looks like an annotation, it is an annotation because - // annotation dictionary fields can be merged with terminal field - // dictionaries. Otherwise, the annotation fields might be there - // to be inherited by annotations below it. + // A dictionary encountered while traversing the /AcroForm field may be a form field, an + // annotation, or the merger of the two. A field that has no fields below it is a terminal. If a + // terminal field looks like an annotation, it is an annotation because annotation dictionary + // fields can be merged with terminal field dictionaries. Otherwise, the annotation fields might + // be there to be inherited by annotations below it. bool is_annotation = false; bool is_field = (0 == depth); @@ -363,8 +352,7 @@ QPDFAcroFormDocumentHelper::traverseField( std::string name = foh.getFullyQualifiedName(); auto old = m->field_to_name.find(f_og); if (old != m->field_to_name.end()) { - // We might be updating after a name change, so remove any - // old information + // We might be updating after a name change, so remove any old information std::string old_name = old->second; m->name_to_fields[old_name].erase(f_og); } @@ -412,11 +400,9 @@ QPDFAcroFormDocumentHelper::generateAppearancesIfNeeded() for (auto& aoh: getWidgetAnnotationsForPage(page)) { QPDFFormFieldObjectHelper ffh = getFieldForAnnotation(aoh); if (ffh.getFieldType() == "/Btn") { - // Rather than generating appearances for button - // fields, rely on what's already there. Just make - // sure /AS is consistent with /V, which we can do by - // resetting the value of the field back to itself. - // This code is referenced in a comment in + // Rather than generating appearances for button fields, rely on what's already + // there. Just make sure /AS is consistent with /V, which we can do by resetting the + // value of the field back to itself. This code is referenced in a comment in // QPDFFormFieldObjectHelper::generateAppearance. if (ffh.isRadioButton() || ffh.isCheckbox()) { ffh.setV(ffh.getValue()); @@ -437,16 +423,13 @@ QPDFAcroFormDocumentHelper::adjustInheritedFields( bool override_q, int from_default_q) { - // Override /Q or /DA if needed. If this object has a field type, - // directly or inherited, it is a field and not just an - // annotation. In that case, we need to override if we are getting - // a value from the document that is different from the value we - // would have gotten from the old document. We must take care not - // to override an explicit value. It's possible that /FT may be - // inherited by lower fields that may explicitly set /DA or /Q or - // that this is a field whose type does not require /DA or /Q and - // we may be put a value on the field that is unused. This is - // harmless, so it's not worth trying to work around. + // Override /Q or /DA if needed. If this object has a field type, directly or inherited, it is a + // field and not just an annotation. In that case, we need to override if we are getting a value + // from the document that is different from the value we would have gotten from the old + // document. We must take care not to override an explicit value. It's possible that /FT may be + // inherited by lower fields that may explicitly set /DA or /Q or that this is a field whose + // type does not require /DA or /Q and we may be put a value on the field that is unused. This + // is harmless, so it's not worth trying to work around. auto has_explicit = [](QPDFFormFieldObjectHelper& field, std::string const& key) { if (field.getObjectHandle().hasKey(key)) { @@ -550,45 +533,36 @@ void QPDFAcroFormDocumentHelper::adjustDefaultAppearances( QPDFObjectHandle obj, std::map> const& dr_map) { - // This method is called on a field that has been copied from - // another file but whose /DA still refers to resources in the - // original file's /DR. - - // When appearance streams are generated for variable text fields - // (see ISO 32000 PDF spec section 12.7.3.3), the field's /DA is - // used to generate content of the appearance stream. /DA contains - // references to resources that may be resolved in the document's - // /DR dictionary, which appears in the document's /AcroForm - // dictionary. For fields that we copied from other documents, we - // need to ensure that resources are mapped correctly in the case - // of conflicting names. For example, if a.pdf's /DR has /F1 - // pointing to one font and b.pdf's /DR also has /F1 but it points - // elsewhere, we need to make sure appearance streams of fields - // copied from b.pdf into a.pdf use whatever font /F1 meant in - // b.pdf, not whatever it means in a.pdf. This method takes care - // of that. It is only called on fields copied from foreign files. + // This method is called on a field that has been copied from another file but whose /DA still + // refers to resources in the original file's /DR. + + // When appearance streams are generated for variable text fields (see ISO 32000 PDF spec + // section 12.7.3.3), the field's /DA is used to generate content of the appearance stream. /DA + // contains references to resources that may be resolved in the document's /DR dictionary, which + // appears in the document's /AcroForm dictionary. For fields that we copied from other + // documents, we need to ensure that resources are mapped correctly in the case of conflicting + // names. For example, if a.pdf's /DR has /F1 pointing to one font and b.pdf's /DR also has /F1 + // but it points elsewhere, we need to make sure appearance streams of fields copied from b.pdf + // into a.pdf use whatever font /F1 meant in b.pdf, not whatever it means in a.pdf. This method + // takes care of that. It is only called on fields copied from foreign files. // A few notes: // - // * If the from document's /DR and the current document's /DR - // have conflicting keys, we have already resolved the conflicts - // before calling this method. The dr_map parameter contains the - // mapping from old keys to new keys. + // * If the from document's /DR and the current document's /DR have conflicting keys, we have + // already resolved the conflicts before calling this method. The dr_map parameter contains + // the mapping from old keys to new keys. // - // * /DA may be inherited from the document's /AcroForm - // dictionary. By the time this method has been called, we have - // already copied any document-level values into the fields to - // avoid having them inherit from the new document. This was - // done in adjustInheritedFields. + // * /DA may be inherited from the document's /AcroForm dictionary. By the time this method has + // been called, we have already copied any document-level values into the fields to avoid + // having them inherit from the new document. This was done in adjustInheritedFields. auto DA = obj.getKey("/DA"); if (!DA.isString()) { return; } - // Find names in /DA. /DA is a string that contains content - // stream-like code, so we create a stream out of the string and - // then filter it. We don't attach the stream to anything, so it + // Find names in /DA. /DA is a string that contains content stream-like code, so we create a + // stream out of the string and then filter it. We don't attach the stream to anything, so it // will get discarded. ResourceFinder rf; auto da_stream = QPDFObjectHandle::newStream(&this->qpdf, DA.getUTF8Value()); @@ -599,8 +573,8 @@ QPDFAcroFormDocumentHelper::adjustDefaultAppearances( QTC::TC("qpdf", "QPDFAcroFormDocumentHelper /DA parse error"); } } catch (std::exception& e) { - // No way to reproduce in test suite right now since error - // conditions are converted to warnings. + // No way to reproduce in test suite right now since error conditions are converted to + // warnings. obj.warnIfPossible( std::string("Unable to parse /DA: ") + e.what() + "; this form field may not update properly"); @@ -620,15 +594,12 @@ void QPDFAcroFormDocumentHelper::adjustAppearanceStream( QPDFObjectHandle stream, std::map> dr_map) { - // We don't have to modify appearance streams or their resource - // dictionaries for them to display properly, but we need to do so - // to make them save to regenerate. Suppose an appearance stream - // as a font /F1 that is different from /F1 in /DR, and that when - // we copy the field, /F1 is remapped to /F1_1. When the field is - // regenerated, /F1_1 won't appear in the stream's resource - // dictionary, so the regenerated appearance stream will revert to - // the /F1_1 in /DR. If we adjust existing appearance streams, we - // are protected from this problem. + // We don't have to modify appearance streams or their resource dictionaries for them to display + // properly, but we need to do so to make them save to regenerate. Suppose an appearance stream + // as a font /F1 that is different from /F1 in /DR, and that when we copy the field, /F1 is + // remapped to /F1_1. When the field is regenerated, /F1_1 won't appear in the stream's resource + // dictionary, so the regenerated appearance stream will revert to the /F1_1 in /DR. If we + // adjust existing appearance streams, we are protected from this problem. auto dict = stream.getDict(); auto resources = dict.getKey("/Resources"); @@ -640,17 +611,15 @@ QPDFAcroFormDocumentHelper::adjustAppearanceStream( resources = this->qpdf.makeIndirectObject(resources); } dict.replaceKey("/Resources", resources); - // Create a dictionary with top-level keys so we can use - // mergeResources to force them to be unshared. We will also use - // this to resolve conflicts that may already be in the resource + // Create a dictionary with top-level keys so we can use mergeResources to force them to be + // unshared. We will also use this to resolve conflicts that may already be in the resource // dictionary. auto merge_with = QPDFObjectHandle::newDictionary(); for (auto const& top_key: dr_map) { merge_with.replaceKey(top_key.first, QPDFObjectHandle::newDictionary()); } resources.mergeResources(merge_with); - // Rename any keys in the resource dictionary that we - // remapped. + // Rename any keys in the resource dictionary that we remapped. for (auto const& i1: dr_map) { std::string const& top_key = i1.first; auto subdict = resources.getKey(top_key); @@ -662,12 +631,10 @@ QPDFAcroFormDocumentHelper::adjustAppearanceStream( std::string const& new_key = i2.second; auto existing_new = subdict.getKey(new_key); if (!existing_new.isNull()) { - // The resource dictionary already has a key in it - // matching what we remapped an old key to, so we'll - // have to move it out of the way. Stick it in - // merge_with, which we will re-merge with the - // dictionary when we're done. We know merge_with - // already has dictionaries for all the top keys. + // The resource dictionary already has a key in it matching what we remapped an old + // key to, so we'll have to move it out of the way. Stick it in merge_with, which we + // will re-merge with the dictionary when we're done. We know merge_with already has + // dictionaries for all the top keys. QTC::TC("qpdf", "QPDFAcroFormDocumentHelper ap conflict"); merge_with.getKey(top_key).replaceKey(new_key, existing_new); } @@ -679,9 +646,8 @@ QPDFAcroFormDocumentHelper::adjustAppearanceStream( } } } - // Deal with any any conflicts by re-merging with merge_with and - // updating our local copy of dr_map, which we will use to modify - // the stream contents. + // Deal with any any conflicts by re-merging with merge_with and updating our local copy of + // dr_map, which we will use to modify the stream contents. resources.mergeResources(merge_with, &dr_map); // Remove empty subdictionaries for (auto iter: resources.ditems()) { @@ -702,8 +668,8 @@ QPDFAcroFormDocumentHelper::adjustAppearanceStream( auto tf = std::shared_ptr(rr); stream.addTokenFilter(tf); } catch (std::exception& e) { - // No way to reproduce in test suite right now since error - // conditions are converted to warnings. + // No way to reproduce in test suite right now since error conditions are converted to + // warnings. stream.warnIfPossible(std::string("Unable to parse appearance stream: ") + e.what()); } } @@ -729,24 +695,22 @@ QPDFAcroFormDocumentHelper::transformAnnotations( } bool foreign = (from_qpdf != &this->qpdf); - // It's possible that we will transform annotations that don't - // include any form fields. This code takes care not to muck - // around with /AcroForm unless we have to. + // It's possible that we will transform annotations that don't include any form fields. This + // code takes care not to muck around with /AcroForm unless we have to. QPDFObjectHandle acroform = this->qpdf.getRoot().getKey("/AcroForm"); QPDFObjectHandle from_acroform = from_qpdf->getRoot().getKey("/AcroForm"); - // /DA and /Q may be inherited from the document-level /AcroForm - // dictionary. If we are copying a foreign stream and the stream - // is getting one of these values from its document's /AcroForm, - // we will need to copy the value explicitly so that it doesn't - // start getting its default from the destination document. + // /DA and /Q may be inherited from the document-level /AcroForm dictionary. If we are copying a + // foreign stream and the stream is getting one of these values from its document's /AcroForm, + // we will need to copy the value explicitly so that it doesn't start getting its default from + // the destination document. bool override_da = false; bool override_q = false; std::string from_default_da; int from_default_q = 0; - // If we copy any form fields, we will need to merge the source - // document's /DR into this document's /DR. + // If we copy any form fields, we will need to merge the source document's /DR into this + // document's /DR. QPDFObjectHandle from_dr = QPDFObjectHandle::newNull(); if (foreign) { std::string default_da; @@ -782,9 +746,8 @@ QPDFAcroFormDocumentHelper::transformAnnotations( } } - // If we have to merge /DR, we will need a mapping of conflicting - // keys for rewriting /DA. Set this up for lazy initialization in - // case we encounter any form fields. + // If we have to merge /DR, we will need a mapping of conflicting keys for rewriting /DA. Set + // this up for lazy initialization in case we encounter any form fields. std::map> dr_map; bool initialized_dr_map = false; QPDFObjectHandle dr = QPDFObjectHandle::newNull(); @@ -804,11 +767,9 @@ QPDFAcroFormDocumentHelper::transformAnnotations( if (!dr.isIndirect()) { dr = acroform.replaceKeyAndGetNew("/DR", this->qpdf.makeIndirectObject(dr)); } - // Merge the other document's /DR, creating a conflict - // map. mergeResources checks to make sure both objects - // are dictionaries. By this point, if this is foreign, - // from_dr has been copied, so we use the target qpdf as - // the owning qpdf. + // Merge the other document's /DR, creating a conflict map. mergeResources checks to + // make sure both objects are dictionaries. By this point, if this is foreign, from_dr + // has been copied, so we use the target qpdf as the owning qpdf. from_dr.makeResourcesIndirect(this->qpdf); dr.mergeResources(from_dr, &dr_map); @@ -818,8 +779,7 @@ QPDFAcroFormDocumentHelper::transformAnnotations( } }; - // This helper prevents us from copying the same object - // multiple times. + // This helper prevents us from copying the same object multiple times. std::map orig_to_copy; auto maybe_copy_object = [&](QPDFObjectHandle& to_copy) { auto og = to_copy.getObjGen(); @@ -842,39 +802,28 @@ QPDFAcroFormDocumentHelper::transformAnnotations( continue; } - // Make copies of annotations and fields down to the - // appearance streams, preserving all internal referential - // integrity. When the incoming annotations are from a - // different file, we first copy them locally. Then, whether - // local or foreign, we copy them again so that if we bring - // the same annotation in multiple times (e.g. overlaying a - // foreign page onto multiple local pages or a local page onto - // multiple other local pages), we don't create annotations - // that are referenced in more than one place. If we did that, - // the effect of applying transformations would be cumulative, - // which is definitely not what we want. Besides, annotations - // and fields are not intended to be referenced in multiple - // places. - - // Determine if this annotation is attached to a form field. - // If so, the annotation may be the same object as the form - // field, or the form field may have the annotation as a kid. - // In either case, we have to walk up the field structure to - // find the top-level field. Within one iteration through a - // set of annotations, we don't want to copy the same item - // more than once. For example, suppose we have field A with - // kids B, C, and D, each of which has annotations BA, CA, and - // DA. When we get to BA, we will find that BA is a kid of B - // which is under A. When we do a copyForeignObject of A, it - // will also copy everything else because of the indirect - // references. When we clone BA, we will want to clone A and - // then update A's clone's kid to point B's clone and B's - // clone's parent to point to A's clone. The same thing holds - // for annotations. Next, when we get to CA, we will again - // discover that A is the top, but we don't want to re-copy A. - // We want CA's clone to be linked to the same clone as BA's. - // Failure to do this will break up things like radio button - // groups, which all have to kids of the same parent. + // Make copies of annotations and fields down to the appearance streams, preserving all + // internal referential integrity. When the incoming annotations are from a different file, + // we first copy them locally. Then, whether local or foreign, we copy them again so that if + // we bring the same annotation in multiple times (e.g. overlaying a foreign page onto + // multiple local pages or a local page onto multiple other local pages), we don't create + // annotations that are referenced in more than one place. If we did that, the effect of + // applying transformations would be cumulative, which is definitely not what we want. + // Besides, annotations and fields are not intended to be referenced in multiple places. + + // Determine if this annotation is attached to a form field. If so, the annotation may be + // the same object as the form field, or the form field may have the annotation as a kid. In + // either case, we have to walk up the field structure to find the top-level field. Within + // one iteration through a set of annotations, we don't want to copy the same item more than + // once. For example, suppose we have field A with kids B, C, and D, each of which has + // annotations BA, CA, and DA. When we get to BA, we will find that BA is a kid of B which + // is under A. When we do a copyForeignObject of A, it will also copy everything else + // because of the indirect references. When we clone BA, we will want to clone A and then + // update A's clone's kid to point B's clone and B's clone's parent to point to A's clone. + // The same thing holds for annotations. Next, when we get to CA, we will again discover + // that A is the top, but we don't want to re-copy A. We want CA's clone to be linked to the + // same clone as BA's. Failure to do this will break up things like radio button groups, + // which all have to kids of the same parent. auto ffield = from_afdh->getFieldForAnnotation(annot); auto ffield_oh = ffield.getObjectHandle(); @@ -886,36 +835,29 @@ QPDFAcroFormDocumentHelper::transformAnnotations( } else if ((!ffield_oh.isNull()) && (!ffield_oh.isIndirect())) { ffield_oh.warnIfPossible("ignoring form field not indirect"); } else if (!ffield_oh.isNull()) { - // A field and its associated annotation can be the same - // object. This matters because we don't want to clone the - // annotation and field separately in this case. + // A field and its associated annotation can be the same object. This matters because we + // don't want to clone the annotation and field separately in this case. have_field = true; // Find the top-level field. It may be the field itself. top_field = ffield.getTopLevelField(&have_parent).getObjectHandle(); if (foreign) { - // copyForeignObject returns the same value if called - // multiple times with the same field. Create/retrieve - // the local copy of the original field. This pulls - // over everything the field references including - // annotations and appearance streams, but it's - // harmless to call copyForeignObject on them too. - // They will already be copied, so we'll get the right - // object back. + // copyForeignObject returns the same value if called multiple times with the same + // field. Create/retrieve the local copy of the original field. This pulls over + // everything the field references including annotations and appearance streams, but + // it's harmless to call copyForeignObject on them too. They will already be copied, + // so we'll get the right object back. // top_field and ffield_oh are known to be indirect. top_field = this->qpdf.copyForeignObject(top_field); ffield_oh = this->qpdf.copyForeignObject(ffield_oh); } else { - // We don't need to add top_field to old_fields if - // it's foreign because the new copy of the foreign - // field won't be referenced anywhere. It's just the - // starting point for us to make an additional local - // copy of. + // We don't need to add top_field to old_fields if it's foreign because the new copy + // of the foreign field won't be referenced anywhere. It's just the starting point + // for us to make an additional local copy of. old_fields.insert(top_field.getObjGen()); } - // Traverse the field, copying kids, and preserving - // integrity. + // Traverse the field, copying kids, and preserving integrity. std::list queue; QPDFObjGen::set seen; if (maybe_copy_object(top_field)) { @@ -933,8 +875,8 @@ QPDFAcroFormDocumentHelper::transformAnnotations( parent.warnIfPossible( "while traversing field " + obj.getObjGen().unparse(',') + ", found parent (" + parent_og.unparse(',') + - ") that had not been seen, indicating likely" - " invalid field structure"); + ") that had not been seen, indicating likely invalid field " + "structure"); } } auto kids = obj.getKey("/Kids"); @@ -955,17 +897,13 @@ QPDFAcroFormDocumentHelper::transformAnnotations( if (foreign) { // Lazily initialize our /DR and the conflict map. init_dr_map(); - // The spec doesn't say anything about /DR on the - // field, but lots of writers put one there, and - // it is frequently the same as the document-level - // /DR. To avoid having the field's /DR point to - // information that we are not maintaining, just - // reset it to that if it exists. Empirical - // evidence suggests that many readers, including - // Acrobat, Adobe Acrobat Reader, chrome, firefox, - // the mac Preview application, and several of the - // free readers on Linux all ignore /DR at the - // field level. + // The spec doesn't say anything about /DR on the field, but lots of writers + // put one there, and it is frequently the same as the document-level /DR. + // To avoid having the field's /DR point to information that we are not + // maintaining, just reset it to that if it exists. Empirical evidence + // suggests that many readers, including Acrobat, Adobe Acrobat Reader, + // chrome, firefox, the mac Preview application, and several of the free + // readers on Linux all ignore /DR at the field level. if (obj.hasKey("/DR")) { obj.replaceKey("/DR", dr); } @@ -1029,8 +967,7 @@ QPDFAcroFormDocumentHelper::transformAnnotations( } } - // Now we can safely mutate the annotation and its appearance - // streams. + // Now we can safely mutate the annotation and its appearance streams. for (auto& stream: streams) { auto dict = stream.getDict(); auto omatrix = dict.getKey("/Matrix"); diff --git a/libqpdf/QPDFFormFieldObjectHelper.cc b/libqpdf/QPDFFormFieldObjectHelper.cc index ef5570f3..67975451 100644 --- a/libqpdf/QPDFFormFieldObjectHelper.cc +++ b/libqpdf/QPDFFormFieldObjectHelper.cc @@ -310,15 +310,15 @@ QPDFFormFieldObjectHelper::setV(QPDFObjectHandle value, bool need_appearances) } } if (!okay) { - this->oh.warnIfPossible("ignoring attempt to set a checkbox field to a" - " value of other than /Yes or /Off"); + this->oh.warnIfPossible("ignoring attempt to set a checkbox field to a value of " + "other than /Yes or /Off"); } } else if (isRadioButton()) { if (value.isName()) { setRadioButtonValue(value); } else { - this->oh.warnIfPossible("ignoring attempt to set a radio button field to" - " an object that is not a name"); + this->oh.warnIfPossible( + "ignoring attempt to set a radio button field to an object that is not a name"); } } else if (isPushbutton()) { this->oh.warnIfPossible("ignoring attempt set the value of a pushbutton field"); @@ -347,24 +347,19 @@ QPDFFormFieldObjectHelper::setV(std::string const& utf8_value, bool need_appeara void QPDFFormFieldObjectHelper::setRadioButtonValue(QPDFObjectHandle name) { - // Set the value of a radio button field. This has the following - // specific behavior: - // * If this is a radio button field that has a parent that is - // also a radio button field and has no explicit /V, call itself - // on the parent - // * If this is a radio button field with children, set /V to the - // given value. Then, for each child, if the child has the - // specified value as one of its keys in the /N subdictionary of - // its /AP (i.e. its normal appearance stream dictionary), set - // /AS to name; otherwise, if /Off is a member, set /AS to /Off. - // Note that we never turn on /NeedAppearances when setting a - // radio button field. + // Set the value of a radio button field. This has the following specific behavior: + // * If this is a radio button field that has a parent that is also a radio button field and has + // no explicit /V, call itself on the parent + // * If this is a radio button field with children, set /V to the given value. Then, for each + // child, if the child has the specified value as one of its keys in the /N subdictionary of + // its /AP (i.e. its normal appearance stream dictionary), set /AS to name; otherwise, if /Off + // is a member, set /AS to /Off. + // Note that we never turn on /NeedAppearances when setting a radio button field. QPDFObjectHandle parent = this->oh.getKey("/Parent"); if (parent.isDictionary() && parent.getKey("/Parent").isNull()) { QPDFFormFieldObjectHelper ph(parent); if (ph.isRadioButton()) { - // This is most likely one of the individual buttons. Try - // calling on the parent. + // This is most likely one of the individual buttons. Try calling on the parent. QTC::TC("qpdf", "QPDFFormFieldObjectHelper set parent radio button"); ph.setRadioButtonValue(name); return; @@ -384,8 +379,7 @@ QPDFFormFieldObjectHelper::setRadioButtonValue(QPDFObjectHandle name) QPDFObjectHandle AP = kid.getKey("/AP"); QPDFObjectHandle annot; if (AP.isNull()) { - // The widget may be below. If there is more than one, - // just find the first one. + // The widget may be below. If there is more than one, just find the first one. QPDFObjectHandle grandkids = kid.getKey("/Kids"); if (grandkids.isArray()) { int ngrandkids = grandkids.getArrayNItems(); @@ -458,9 +452,8 @@ void QPDFFormFieldObjectHelper::generateAppearance(QPDFAnnotationObjectHelper& aoh) { std::string ft = getFieldType(); - // Ignore field types we don't know how to generate appearances - // for. Button fields don't really need them -- see code in - // QPDFAcroFormDocumentHelper::generateAppearancesIfNeeded. + // Ignore field types we don't know how to generate appearances for. Button fields don't really + // need them -- see code in QPDFAcroFormDocumentHelper::generateAppearancesIfNeeded. if ((ft == "/Tx") || (ft == "/Ch")) { generateTextAppearance(aoh); } @@ -562,15 +555,13 @@ ValueSetter::writeAppearance() { this->replaced = true; - // This code does not take quadding into consideration because - // doing so requires font metric information, which we don't - // have in many cases. + // This code does not take quadding into consideration because doing so requires font metric + // information, which we don't have in many cases. double tfh = 1.2 * tf; int dx = 1; - // Write one or more lines, centered vertically, possibly with - // one row highlighted. + // Write one or more lines, centered vertically, possibly with one row highlighted. auto max_rows = static_cast((bbox.ury - bbox.lly) / tfh); bool highlight = false; @@ -591,8 +582,7 @@ ValueSetter::writeAppearance() } } if (found) { - // Try to make the found item the second one, but - // adjust for under/overflow. + // Try to make the found item the second one, but adjust for under/overflow. int wanted_first = QIntC::to_int(found_idx) - 1; int wanted_last = QIntC::to_int(found_idx + max_rows) - 2; QTC::TC("qpdf", "QPDFFormFieldObjectHelper list found"); @@ -639,9 +629,8 @@ ValueSetter::writeAppearance() dy -= tf; write("q\nBT\n" + DA + "\n"); for (size_t i = 0; i < nlines; ++i) { - // We could adjust Tm to translate to the beginning the first - // line, set TL to tfh, and use T* for each subsequent line, - // but doing this would require extracting any Tm from DA, + // We could adjust Tm to translate to the beginning the first line, set TL to tfh, and use + // T* for each subsequent line, but doing this would require extracting any Tm from DA, // which doesn't seem really worth the effort. if (i == 0) { write( @@ -708,8 +697,8 @@ TfFinder::handleToken(QPDFTokenizer::Token const& token) case QPDFTokenizer::tt_word: if (token.isWord("Tf")) { if ((last_num > 1.0) && (last_num < 1000.0)) { - // These ranges are arbitrary but keep us from doing - // insane things or suffering from over/underflow + // These ranges are arbitrary but keep us from doing insane things or suffering from + // over/underflow tf = last_num; } tf_idx = last_num_idx; @@ -738,8 +727,7 @@ TfFinder::getDA() if (QIntC::to_int(i) == tf_idx) { double delta = strtod(cur.c_str(), nullptr) - this->tf; if ((delta > 0.001) || (delta < -0.001)) { - // tf doesn't match the font size passed to Tf, so - // substitute. + // tf doesn't match the font size passed to Tf, so substitute. QTC::TC("qpdf", "QPDFFormFieldObjectHelper fallback Tf"); cur = QUtil::double_to_string(tf); } @@ -852,6 +840,5 @@ QPDFFormFieldObjectHelper::generateTextAppearance(QPDFAnnotationObjectHelper& ao } AS.addTokenFilter( - // line-break std::shared_ptr(new ValueSetter(DA, V, opt, tf, bbox))); } diff --git a/libqpdf/QPDFJob.cc b/libqpdf/QPDFJob.cc index 50ea5ea7..4c7fdd04 100644 --- a/libqpdf/QPDFJob.cc +++ b/libqpdf/QPDFJob.cc @@ -130,8 +130,8 @@ ImageOptimizer::makePipeline(std::string const& description, Pipeline* next) if (!(w_obj.isNumber() && h_obj.isNumber())) { if (!description.empty()) { o.doIfVerbose([&](Pipeline& v, std::string const& prefix) { - v << prefix << ": " << description << ": not optimizing because image dictionary" - << " is missing required keys\n"; + v << prefix << ": " << description + << ": not optimizing because image dictionary is missing required keys\n"; }); } return result; @@ -142,14 +142,13 @@ ImageOptimizer::makePipeline(std::string const& description, Pipeline* next) if (!description.empty()) { o.doIfVerbose([&](Pipeline& v, std::string const& prefix) { v << prefix << ": " << description - << ": not optimizing because image has other than" - << " 8 bits per component\n"; + << ": not optimizing because image has other than 8 bits per component\n"; }); } return result; } - // Files have been seen in the wild whose width and height are - // floating point, which is goofy, but we can deal with it. + // Files have been seen in the wild whose width and height are floating point, which is goofy, + // but we can deal with it. JDIMENSION w = 0; if (w_obj.isInteger()) { w = w_obj.getUIntValueAsUInt(); @@ -178,8 +177,8 @@ ImageOptimizer::makePipeline(std::string const& description, Pipeline* next) QTC::TC("qpdf", "QPDFJob image optimize colorspace"); if (!description.empty()) { o.doIfVerbose([&](Pipeline& v, std::string const& prefix) { - v << prefix << ": " << description << ": not optimizing because qpdf can't optimize" - << " images with this colorspace\n"; + v << prefix << ": " << description + << ": not optimizing because qpdf can't optimize images with this colorspace\n"; }); } return result; @@ -190,8 +189,9 @@ ImageOptimizer::makePipeline(std::string const& description, Pipeline* next) QTC::TC("qpdf", "QPDFJob image optimize too small"); if (!description.empty()) { o.doIfVerbose([&](Pipeline& v, std::string const& prefix) { - v << prefix << ": " << description << ": not optimizing because image" - << " is smaller than requested minimum dimensions\n"; + v << prefix << ": " << description + << ": not optimizing because image is smaller than requested minimum " + "dimensions\n"; }); } return result; @@ -207,8 +207,8 @@ ImageOptimizer::evaluate(std::string const& description) if (!image.pipeStreamData(nullptr, 0, qpdf_dl_specialized, true)) { QTC::TC("qpdf", "QPDFJob image optimize no pipeline"); o.doIfVerbose([&](Pipeline& v, std::string const& prefix) { - v << prefix << ": " << description << ": not optimizing because unable to decode data" - << " or data already uses DCT\n"; + v << prefix << ": " << description + << ": not optimizing because unable to decode data or data already uses DCT\n"; }); return false; } @@ -227,8 +227,7 @@ ImageOptimizer::evaluate(std::string const& description) QTC::TC("qpdf", "QPDFJob image optimize no shrink"); o.doIfVerbose([&](Pipeline& v, std::string const& prefix) { v << prefix << ": " << description - << ": not optimizing because DCT compression does not" - << " reduce image size\n"; + << ": not optimizing because DCT compression does not reduce image size\n"; }); return false; } @@ -245,8 +244,8 @@ ImageOptimizer::provideStreamData(QPDFObjGen const&, Pipeline* pipeline) std::shared_ptr p = makePipeline("", pipeline); if (p == nullptr) { // Should not be possible - image.warnIfPossible("unable to create pipeline after previous" - " success; image data will be lost"); + image.warnIfPossible( + "unable to create pipeline after previous success; image data will be lost"); pipeline->finish(); return; } @@ -441,8 +440,7 @@ QPDFJob::createQPDF() processFile(pdf_sp, m->infilename.get(), m->password.get(), true, true); } catch (QPDFExc& e) { if (e.getErrorCode() == qpdf_e_password) { - // Allow certain operations to work when an incorrect - // password is supplied. + // Allow certain operations to work when an incorrect password is supplied. if (m->check_is_encrypted || m->check_requires_password) { m->encryption_status = qpdf_es_encrypted | qpdf_es_password_incorrect; return nullptr; @@ -464,8 +462,8 @@ QPDFJob::createQPDF() return nullptr; } - // If we are updating from JSON, this has to be done first before - // other options may cause transformations to the input. + // If we are updating from JSON, this has to be done first before other options may cause + // transformations to the input. if (!m->update_from_json.empty()) { pdf.updateFromJSON(m->update_from_json); } @@ -497,16 +495,16 @@ QPDFJob::writeQPDF(QPDF& pdf) } if (m->warnings && (!m->suppress_warnings)) { if (createsOutput()) { - *m->log->getWarn() << m->message_prefix << ": operation succeeded with warnings;" - << " resulting file may have some problems\n"; + *m->log->getWarn() + << m->message_prefix + << ": operation succeeded with warnings; resulting file may have some problems\n"; } else { *m->log->getWarn() << m->message_prefix << ": operation succeeded with warnings\n"; } } if (m->report_mem_usage) { - // Call get_max_memory_usage before generating output. When - // debugging, it's easier if print statements from - // get_max_memory_usage are not interleaved with the output. + // Call get_max_memory_usage before generating output. When debugging, it's easier if print + // statements from get_max_memory_usage are not interleaved with the output. auto mem_usage = QUtil::get_max_memory_usage(); *m->log->getWarn() << "qpdf-max-memory-usage " << mem_usage << "\n"; } @@ -568,16 +566,13 @@ QPDFJob::getExitCode() const void QPDFJob::checkConfiguration() { - // Do final checks for command-line consistency. (I always think - // this is called doFinalChecks, so I'm putting that in a - // comment.) + // Do final checks for command-line consistency. (I always think this is called doFinalChecks, + // so I'm putting that in a comment.) if (m->replace_input) { - // Check for --empty appears later after we have checked - // m->infilename. + // Check for --empty appears later after we have checked m->infilename. if (m->outfilename) { - usage("--replace-input may not be used when" - " an output file is specified"); + usage("--replace-input may not be used when an output file is specified"); } else if (m->split_pages) { usage("--split-pages may not be used with --replace-input"); } else if (m->json_version) { @@ -585,8 +580,8 @@ QPDFJob::checkConfiguration() } } if (m->json_version && (m->outfilename == nullptr)) { - // The output file is optional with --json for backward - // compatibility and defaults to standard output. + // The output file is optional with --json for backward compatibility and defaults to + // standard output. m->outfilename = QUtil::make_shared_cstr("-"); } if (m->infilename == nullptr) { @@ -605,24 +600,21 @@ QPDFJob::checkConfiguration() if (m->encrypt && (!m->allow_insecure) && (m->owner_password.empty() && (!m->user_password.empty()) && (m->keylen == 256))) { - // Note that empty owner passwords for R < 5 are copied from - // the user password, so this lack of security is not an issue - // for those files. Also we are consider only the ability to - // open the file without a password to be insecure. We are not - // concerned about whether the viewer enforces security - // settings when the user and owner password match. - usage("A PDF with a non-empty user password and an empty owner" - " password encrypted with a 256-bit key is insecure as it" - " can be opened without a password. If you really want to" - " do this, you must also give the --allow-insecure option" - " before the -- that follows --encrypt."); + // Note that empty owner passwords for R < 5 are copied from the user password, so this lack + // of security is not an issue for those files. Also we are consider only the ability to + // open the file without a password to be insecure. We are not concerned about whether the + // viewer enforces security settings when the user and owner password match. + usage( + "A PDF with a non-empty user password and an empty owner password encrypted with a " + "256-bit key is insecure as it can be opened without a password. If you really want to" + " do this, you must also give the --allow-insecure option before the -- that follows " + "--encrypt."); } bool save_to_stdout = false; if (m->require_outfile && m->outfilename && (strcmp(m->outfilename.get(), "-") == 0)) { if (m->split_pages) { - usage("--split-pages may not be used when" - " writing to standard output"); + usage("--split-pages may not be used when writing to standard output"); } save_to_stdout = true; } @@ -634,9 +626,8 @@ QPDFJob::checkConfiguration() } if ((!m->split_pages) && QUtil::same_file(m->infilename.get(), m->outfilename.get())) { QTC::TC("qpdf", "QPDFJob same file error"); - usage("input file and output file are the same;" - " use --replace-input to intentionally" - " overwrite the input file"); + usage("input file and output file are the same; use --replace-input to intentionally " + "overwrite the input file"); } if (m->json_version == 1) { @@ -645,8 +636,7 @@ QPDFJob::checkConfiguration() } } else { if (m->json_keys.count("objectinfo") || m->json_keys.count("objects")) { - usage("json keys \"objects\" and \"objectinfo\" are only valid for" - " json version 1"); + usage("json keys \"objects\" and \"objectinfo\" are only valid for json version 1"); } } } @@ -754,10 +744,8 @@ QPDFJob::showEncryption(QPDF& pdf) void QPDFJob::doCheck(QPDF& pdf) { - // Code below may set okay to false but not to true. - // We assume okay until we prove otherwise but may - // continue to perform additional checks after finding - // errors. + // Code below may set okay to false but not to true. We assume okay until we prove otherwise but + // may continue to perform additional checks after finding errors. bool okay = true; auto& cout = *m->log->getInfo(); cout << "checking " << m->infilename.get() << "\n"; @@ -777,8 +765,7 @@ QPDFJob::doCheck(QPDF& pdf) cout << "File is not linearized\n"; } - // Write the file to nowhere, uncompressing - // streams. This causes full file traversal and + // Write the file to nowhere, uncompressing streams. This causes full file traversal and // decoding of all streams we can decode. QPDFWriter w(pdf); Pl_Discard discard; @@ -809,9 +796,9 @@ QPDFJob::doCheck(QPDF& pdf) if (!pdf.getWarnings().empty()) { m->warnings = true; } else { - *m->log->getInfo() << "No syntax or stream encoding errors" - << " found; the file may still contain\n" - << "errors that qpdf cannot detect\n"; + *m->log->getInfo() + << "No syntax or stream encoding errors found; the file may still contain\n" + << "errors that qpdf cannot detect\n"; } } @@ -833,8 +820,7 @@ QPDFJob::doShowObj(QPDF& pdf) obj.warnIfPossible("unable to filter stream data"); error = true; } else { - // If anything has been written to standard output, - // this will fail. + // If anything has been written to standard output, this will fail. m->log->saveToStandardOutput(true); obj.pipeStreamData( m->log->getSave().get(), @@ -933,8 +919,8 @@ QPDFJob::doShowAttachment(QPDF& pdf) throw std::runtime_error("attachment " + m->attachment_to_show + " not found"); } auto efs = fs->getEmbeddedFileStream(); - // saveToStandardOutput has already been called, but it's harmless - // to call it again, so do as defensive coding. + // saveToStandardOutput has already been called, but it's harmless to call it again, so do as + // defensive coding. m->log->saveToStandardOutput(true); efs.pipeStreamData(m->log->getSave().get(), 0, qpdf_dl_all); } @@ -1132,9 +1118,8 @@ QPDFJob::doJSONPageLabels(Pipeline* p, bool& first, QPDF& pdf) pldh.getLabelsForPageRange(0, npages - 1, 0, labels); for (auto iter = labels.begin(); iter != labels.end(); ++iter) { if ((iter + 1) == labels.end()) { - // This can't happen, so ignore it. This could only - // happen if getLabelsForPageRange somehow returned an - // odd number of items. + // This can't happen, so ignore it. This could only happen if getLabelsForPageRange + // somehow returned an odd number of items. break; } JSON j_label = j_labels.addArrayElement(JSON::makeDictionary()); @@ -1362,22 +1347,17 @@ QPDFJob::doJSONAttachments(Pipeline* p, bool& first, QPDF& pdf) JSON QPDFJob::json_schema(int json_version, std::set* keys) { - // Style: use all lower-case keys with no dashes or underscores. - // Choose array or dictionary based on indexing. For example, we - // use a dictionary for objects because we want to index by object - // ID and an array for pages because we want to index by position. - // The pages in the pages array contain references back to the - // original object, which can be resolved in the objects - // dictionary. When a PDF construct that maps back to an original - // object is represented separately, use "object" as the key that - // references the original object. - - // This JSON object doubles as a schema and as documentation for - // our JSON output. Any schema mismatch is a bug in qpdf. This - // helps to enforce our policy of consistently providing a known - // structure where every documented key will always be present, - // which makes it easier to consume our JSON. This is discussed in - // more depth in the manual. + // Style: use all lower-case keys with no dashes or underscores. Choose array or dictionary + // based on indexing. For example, we use a dictionary for objects because we want to index by + // object ID and an array for pages because we want to index by position. The pages in the pages + // array contain references back to the original object, which can be resolved in the objects + // dictionary. When a PDF construct that maps back to an original object is represented + // separately, use "object" as the key that references the original object. + + // This JSON object doubles as a schema and as documentation for our JSON output. Any schema + // mismatch is a bug in qpdf. This helps to enforce our policy of consistently providing a known + // structure where every documented key will always be present, which makes it easier to consume + // our JSON. This is discussed in more depth in the manual. JSON schema = JSON::makeDictionary(); schema.addDictionaryMember( "version", @@ -1388,9 +1368,8 @@ QPDFJob::json_schema(int json_version, std::set* keys) bool all_keys = ((keys == nullptr) || keys->empty()); - // The list of selectable top-level keys id duplicated in the - // following places: job.yml, QPDFJob::json_schema, and - // QPDFJob::doJSON. + // The list of selectable top-level keys id duplicated in the following places: job.yml, + // QPDFJob::json_schema, and QPDFJob::doJSON. if (json_version == 1) { if (all_keys || keys->count("objects")) { schema.addDictionaryMember("objects", JSON::parse(R"({ @@ -1581,8 +1560,8 @@ QPDFJob::json_out_schema_v1() void QPDFJob::doJSON(QPDF& pdf, Pipeline* p) { - // qpdf guarantees that no new top-level keys whose names start - // with "x-" will be added. These are reserved for users. + // qpdf guarantees that no new top-level keys whose names start with "x-" will be added. These + // are reserved for users. std::string captured_json; std::shared_ptr pl_str; @@ -1595,14 +1574,12 @@ QPDFJob::doJSON(QPDF& pdf, Pipeline* p) JSON::writeDictionaryOpen(p, first, 0); if (m->json_output) { - // Exclude version and parameters to keep the output file - // minimal. The JSON version is inside the "qpdf" key for - // version 2. + // Exclude version and parameters to keep the output file minimal. The JSON version is + // inside the "qpdf" key for version 2. } else { - // This version is updated every time a non-backward-compatible - // change is made to the JSON format. Clients of the JSON are to - // ignore unrecognized keys, so we only update the version of a - // key disappears or if its value changes meaning. + // This version is updated every time a non-backward-compatible change is made to the JSON + // format. Clients of the JSON are to ignore unrecognized keys, so we only update the + // version of a key disappears or if its value changes meaning. JSON::writeDictionaryItem(p, first, "version", JSON::makeInt(m->json_version), 1); JSON j_params = JSON::makeDictionary(); std::string decode_level_str; @@ -1624,13 +1601,11 @@ QPDFJob::doJSON(QPDF& pdf, Pipeline* p) JSON::writeDictionaryItem(p, first, "parameters", j_params, 1); } bool all_keys = m->json_keys.empty(); - // The list of selectable top-level keys id duplicated in the - // following places: job.yml, QPDFJob::json_schema, and - // QPDFJob::doJSON. + // The list of selectable top-level keys id duplicated in the following places: job.yml, + // QPDFJob::json_schema, and QPDFJob::doJSON. - // We do pages and pagelabels first since they have the side - // effect of repairing the pages tree, which could potentially - // impact object references in remaining items. + // We do pages and pagelabels first since they have the side effect of repairing the pages tree, + // which could potentially impact object references in remaining items. if (all_keys || m->json_keys.count("pages")) { doJSONPages(p, first, pdf); } @@ -1638,8 +1613,7 @@ QPDFJob::doJSON(QPDF& pdf, Pipeline* p) doJSONPageLabels(p, first, pdf); } - // The non-special keys are output in alphabetical order, but the - // order doesn't actually matter. + // The non-special keys are output in alphabetical order, but the order doesn't actually matter. if (all_keys || m->json_keys.count("acroform")) { doJSONAcroform(p, first, pdf); } @@ -1653,16 +1627,15 @@ QPDFJob::doJSON(QPDF& pdf, Pipeline* p) doJSONOutlines(p, first, pdf); } - // We do objects last so their information is consistent with - // repairing the page tree. To see the original file with any page - // tree problems and the page tree not flattened, select + // We do objects last so their information is consistent with repairing the page tree. To see + // the original file with any page tree problems and the page tree not flattened, select // qpdf/objects/objectinfo without other keys. if (all_keys || m->json_keys.count("objects") || m->json_keys.count("qpdf")) { doJSONObjects(p, first, pdf); } if (m->json_version == 1) { - // "objectinfo" is not needed for version >1 since you can - // tell streams from other objects in "objects". + // "objectinfo" is not needed for version >1 since you can tell streams from other objects + // in "objects". if (all_keys || m->json_keys.count("objectinfo")) { doJSONObjectinfo(p, first, pdf); } @@ -1677,8 +1650,7 @@ QPDFJob::doJSON(QPDF& pdf, Pipeline* p) std::list errors; JSON captured = JSON::parse(captured_json); if (!captured.checkSchema(schema, errors)) { - m->log->error("QPDFJob didn't create JSON that complies with " - "its own rules.\n"); + m->log->error("QPDFJob didn't create JSON that complies with its own rules.\n"); for (auto const& error: errors) { *m->log->getError() << error << "\n"; } @@ -1768,53 +1740,46 @@ QPDFJob::doProcess( bool used_for_input, bool main_input) { - // If a password has been specified but doesn't work, try other - // passwords that are equivalent in different character encodings. - // This makes it possible to open PDF files that were encrypted - // using incorrect string encodings. For example, if someone used - // a password encoded in PDF Doc encoding or Windows code page - // 1252 for an AES-encrypted file or a UTF-8-encoded password on - // an RC4-encrypted file, or if the password was properly encoded - // but the password given here was incorrectly encoded, there's a - // good chance we'd succeed here. + // If a password has been specified but doesn't work, try other passwords that are equivalent in + // different character encodings. This makes it possible to open PDF files that were encrypted + // using incorrect string encodings. For example, if someone used a password encoded in PDF Doc + // encoding or Windows code page 1252 for an AES-encrypted file or a UTF-8-encoded password on + // an RC4-encrypted file, or if the password was properly encoded but the password given here + // was incorrectly encoded, there's a good chance we'd succeed here. std::string ptemp; if (password && (!m->password_is_hex_key)) { if (m->password_mode == QPDFJob::pm_hex_bytes) { - // Special case: handle --password-mode=hex-bytes for input - // password as well as output password + // Special case: handle --password-mode=hex-bytes for input password as well as output + // password QTC::TC("qpdf", "QPDFJob input password hex-bytes"); ptemp = QUtil::hex_decode(password); password = ptemp.c_str(); } } if ((password == nullptr) || empty || m->password_is_hex_key || m->suppress_password_recovery) { - // There is no password, or we're not doing recovery, so just - // do the normal processing with the supplied password. + // There is no password, or we're not doing recovery, so just do the normal processing with + // the supplied password. doProcessOnce(pdf, fn, password, empty, used_for_input, main_input); return; } - // Get a list of otherwise encoded strings. Keep in scope for this - // method. + // Get a list of otherwise encoded strings. Keep in scope for this method. std::vector passwords_str = QUtil::possible_repaired_encodings(password); // Represent to char const*, as required by the QPDF class. std::vector passwords; for (auto const& iter: passwords_str) { passwords.push_back(iter.c_str()); } - // We always try the supplied password first because it is the - // first string returned by possible_repaired_encodings. If there - // is more than one option, go ahead and put the supplied password - // at the end so that it's that decoding attempt whose exception - // is thrown. + // We always try the supplied password first because it is the first string returned by + // possible_repaired_encodings. If there is more than one option, go ahead and put the supplied + // password at the end so that it's that decoding attempt whose exception is thrown. if (passwords.size() > 1) { passwords.push_back(password); } - // Try each password. If one works, return the resulting object. - // If they all fail, throw the exception thrown by the final - // attempt, which, like the first attempt, will be with the + // Try each password. If one works, return the resulting object. If they all fail, throw the + // exception thrown by the final attempt, which, like the first attempt, will be with the // supplied password. bool warned = false; for (auto iter = passwords.begin(); iter != passwords.end(); ++iter) { @@ -1831,9 +1796,9 @@ QPDFJob::doProcess( if (!warned) { warned = true; doIfVerbose([&](Pipeline& v, std::string const& prefix) { - v << prefix << ": supplied password didn't work;" - << " trying other passwords based on interpreting" - << " password with different string encodings\n"; + v << prefix + << ": supplied password didn't work; trying other passwords based on " + "interpreting password with different string encodings\n"; }); } } @@ -1943,10 +1908,8 @@ QPDFJob::doUnderOverlayForPage( fo[from_pageno] = pdf.copyForeignObject(from_page.getFormXObjectForPage()); } - // If the same page is overlaid or underlaid multiple times, - // we'll generate multiple names for it, but that's harmless - // and also a pretty goofy case that's not worth coding - // around. + // If the same page is overlaid or underlaid multiple times, we'll generate multiple names + // for it, but that's harmless and also a pretty goofy case that's not worth coding around. std::string name = resources.getUniqueResourceName("/Fx", min_suffix); QPDFMatrix cm; std::string new_content = dest_page.placeFormXObject( @@ -2017,18 +1980,15 @@ QPDFJob::handleUnderOverlay(QPDF& pdf) if (!(underlay_pagenos.count(pageno) || overlay_pagenos.count(pageno))) { continue; } - // This code converts the original page, any underlays, and - // any overlays to form XObjects. Then it concatenates display - // of all underlays, the original page, and all overlays. - // Prior to 11.3.0, the original page contents were wrapped in - // q/Q, but this didn't work if the original page had - // unbalanced q/Q operators. See github issue #904. + // This code converts the original page, any underlays, and any overlays to form XObjects. + // Then it concatenates display of all underlays, the original page, and all overlays. Prior + // to 11.3.0, the original page contents were wrapped in q/Q, but this didn't work if the + // original page had unbalanced q/Q operators. See github issue #904. auto& dest_page = main_pages.at(i); auto dest_page_oh = dest_page.getObjectHandle(); auto this_page_fo = dest_page.getFormXObjectForPage(); - // The resulting form xobject lazily reads the content from - // the original page, which we are going to replace. Therefore - // we have to explicitly copy it. + // The resulting form xobject lazily reads the content from the original page, which we are + // going to replace. Therefore we have to explicitly copy it. auto content_data = this_page_fo.getRawStreamData(); this_page_fo.replaceStreamData(content_data, QPDFObjectHandle(), QPDFObjectHandle()); auto resources = @@ -2097,8 +2057,7 @@ QPDFJob::addAttachments(QPDF& pdf) } message = pdf.getFilename() + " already has attachments with the following keys: " + message + - "; use --replace to replace or --key to specify a different " - "key"; + "; use --replace to replace or --key to specify a different key"; throw std::runtime_error(message); } } @@ -2144,11 +2103,9 @@ QPDFJob::copyAttachments(QPDF& pdf) message += i; } message = pdf.getFilename() + - " already has attachments with keys that conflict with" - " attachments from other files: " + + " already has attachments with keys that conflict with attachments from other files: " + message + - ". Use --prefix with --copy-attachments-from" - " or manually copy individual attachments."; + ". Use --prefix with --copy-attachments-from or manually copy individual attachments."; throw std::runtime_error(message); } } @@ -2243,13 +2200,11 @@ QPDFJob::shouldRemoveUnreferencedResources(QPDF& pdf) return true; } - // Unreferenced resources are common in files where resources - // dictionaries are shared across pages. As a heuristic, we look - // in the file for shared resources dictionaries or shared XObject - // subkeys of resources dictionaries either on pages or on form - // XObjects in pages. If we find any, then there is a higher - // likelihood that the expensive process of finding unreferenced - // resources is worth it. + // Unreferenced resources are common in files where resources dictionaries are shared across + // pages. As a heuristic, we look in the file for shared resources dictionaries or shared + // XObject subkeys of resources dictionaries either on pages or on form XObjects in pages. If we + // find any, then there is a higher likelihood that the expensive process of finding + // unreferenced resources is worth it. // Return true as soon as we find any shared resources. @@ -2332,8 +2287,8 @@ added_page(QPDF& pdf, QPDFObjectHandle page) { QPDFObjectHandle result = page; if (&page.getQPDF() != &pdf) { - // Calling copyForeignObject on an object we already copied - // will give us the already existing copy. + // Calling copyForeignObject on an object we already copied will give us the already + // existing copy. result = pdf.copyForeignObject(page); } return result; @@ -2348,8 +2303,7 @@ added_page(QPDF& pdf, QPDFPageObjectHelper page) void QPDFJob::handlePageSpecs(QPDF& pdf, std::vector>& page_heap) { - // Parse all page specifications and translate them into lists of - // actual pages. + // Parse all page specifications and translate them into lists of actual pages. // Handle "." as a shortcut for the input file for (auto& page_spec: m->page_specs) { @@ -2359,9 +2313,8 @@ QPDFJob::handlePageSpecs(QPDF& pdf, std::vector>& page_hea } if (!m->keep_files_open_set) { - // Count the number of distinct files to determine whether we - // should keep files open or not. Rather than trying to code - // some portable heuristic based on OS limits, just hard-code + // Count the number of distinct files to determine whether we should keep files open or not. + // Rather than trying to code some portable heuristic based on OS limits, just hard-code // this at a given number and allow users to override. std::set filenames; for (auto& page_spec: m->page_specs) { @@ -2383,16 +2336,13 @@ QPDFJob::handlePageSpecs(QPDF& pdf, std::vector>& page_hea std::map> copied_pages; for (auto& page_spec: m->page_specs) { if (page_spec_qpdfs.count(page_spec.filename) == 0) { - // Open the PDF file and store the QPDF object. Throw a - // std::shared_ptr to the qpdf into a heap so that it - // survives through copying to the output but gets cleaned up - // automatically at the end. Do not canonicalize the file - // name. Using two different paths to refer to the same - // file is a documented workaround for duplicating a page. - // If you are using this an example of how to do this with - // the API, you can just create two different QPDF objects - // to the same underlying file with the same path to - // achieve the same affect. + // Open the PDF file and store the QPDF object. Throw a std::shared_ptr to the qpdf into + // a heap so that it survives through copying to the output but gets cleaned up + // automatically at the end. Do not canonicalize the file name. Using two different + // paths to refer to the same file is a documented workaround for duplicating a page. If + // you are using this an example of how to do this with the API, you can just create two + // different QPDF objects to the same underlying file with the same path to achieve the + // same affect. char const* password = page_spec.password.get(); if ((!m->encryption_file.empty()) && (password == nullptr) && (page_spec.filename == m->encryption_file)) { @@ -2424,8 +2374,8 @@ QPDFJob::handlePageSpecs(QPDF& pdf, std::vector>& page_hea } } - // Read original pages from the PDF, and parse the page range - // associated with this occurrence of the file. + // Read original pages from the PDF, and parse the page range associated with this + // occurrence of the file. parsed_specs.push_back( // line-break QPDFPageData(page_spec.filename, page_spec_qpdfs[page_spec.filename], page_spec.range)); @@ -2451,11 +2401,9 @@ QPDFJob::handlePageSpecs(QPDF& pdf, std::vector>& page_hea } } - // Clear all pages out of the primary QPDF's pages tree but leave - // the objects in place in the file so they can be re-added - // without changing their object numbers. This enables other - // things in the original file, such as outlines, to continue to - // work. + // Clear all pages out of the primary QPDF's pages tree but leave the objects in place in the + // file so they can be re-added without changing their object numbers. This enables other things + // in the original file, such as outlines, to continue to work. doIfVerbose([&](Pipeline& v, std::string const& prefix) { v << prefix << ": removing unreferenced pages from primary input\n"; }); @@ -2466,9 +2414,8 @@ QPDFJob::handlePageSpecs(QPDF& pdf, std::vector>& page_hea } if (m->collate && (parsed_specs.size() > 1)) { - // Collate the pages by selecting one page from each spec in - // order. When a spec runs out of pages, stop selecting from - // it. + // Collate the pages by selecting one page from each spec in order. When a spec runs out of + // pages, stop selecting from it. std::vector new_parsed_specs; size_t nspecs = parsed_specs.size(); size_t cur_page = 0; @@ -2491,9 +2438,8 @@ QPDFJob::handlePageSpecs(QPDF& pdf, std::vector>& page_hea parsed_specs = new_parsed_specs; } - // Add all the pages from all the files in the order specified. - // Keep track of any pages from the original file that we are - // selecting. + // Add all the pages from all the files in the order specified. Keep track of any pages from the + // original file that we are selecting. std::set selected_from_orig; std::vector new_labels; bool any_page_labels = false; @@ -2516,8 +2462,7 @@ QPDFJob::handlePageSpecs(QPDF& pdf, std::vector>& page_hea v << prefix << ": adding pages from " << page_data.filename << "\n"; }); for (auto pageno_iter: page_data.selected_pages) { - // Pages are specified from 1 but numbered from 0 in the - // vector + // Pages are specified from 1 but numbered from 0 in the vector int pageno = pageno_iter - 1; pldh.getLabelsForPageRange(pageno, pageno, out_pageno++, new_labels); QPDFPageObjectHelper to_copy = page_data.orig_pages.at(QIntC::to_size(pageno)); @@ -2539,22 +2484,18 @@ QPDFJob::handlePageSpecs(QPDF& pdf, std::vector>& page_hea bool first_copy_from_orig = false; bool this_file = (page_data.qpdf == &pdf); if (this_file) { - // This is a page from the original file. Keep track - // of the fact that we are using it. + // This is a page from the original file. Keep track of the fact that we are using + // it. first_copy_from_orig = (selected_from_orig.count(pageno) == 0); selected_from_orig.insert(pageno); } auto new_page = added_page(pdf, to_copy); - // Try to avoid gratuitously renaming fields. In the case - // of where we're just extracting a bunch of pages from - // the original file and not copying any page more than - // once, there's no reason to do anything with the fields. - // Since we don't remove fields from the original file - // until all copy operations are completed, any foreign - // pages that conflict with original pages will be - // adjusted. If we copy any page from the original file - // more than once, that page would be in conflict with the - // previous copy of itself. + // Try to avoid gratuitously renaming fields. In the case of where we're just extracting + // a bunch of pages from the original file and not copying any page more than once, + // there's no reason to do anything with the fields. Since we don't remove fields from + // the original file until all copy operations are completed, any foreign pages that + // conflict with original pages will be adjusted. If we copy any page from the original + // file more than once, that page would be in conflict with the previous copy of itself. if (other_afdh->hasAcroForm() && ((!this_file) || (!first_copy_from_orig))) { if (!this_file) { QTC::TC("qpdf", "QPDFJob copy fields not this file"); @@ -2569,8 +2510,8 @@ QPDFJob::handlePageSpecs(QPDF& pdf, std::vector>& page_hea qpdf_e_damaged_pdf, "", 0, - ("Exception caught while fixing copied" - " annotations. This may be a qpdf bug. " + + ("Exception caught while fixing copied annotations. This may be a qpdf " + "bug. " + std::string("Exception: ") + e.what())); } } @@ -2585,10 +2526,9 @@ QPDFJob::handlePageSpecs(QPDF& pdf, std::vector>& page_hea pdf.getRoot().replaceKey("/PageLabels", page_labels); } - // Delete page objects for unused page in primary. This prevents - // those objects from being preserved by being referred to from - // other places, such as the outlines dictionary. Also make sure - // we keep form fields from pages we preserved. + // Delete page objects for unused page in primary. This prevents those objects from being + // preserved by being referred to from other places, such as the outlines dictionary. Also make + // sure we keep form fields from pages we preserved. for (size_t pageno = 0; pageno < orig_pages.size(); ++pageno) { auto page = orig_pages.at(pageno); if (selected_from_orig.count(QIntC::to_int(pageno))) { @@ -2676,8 +2616,8 @@ QPDFJob::maybeFixWritePassword(int R, std::string& password) std::string encoded; if (!QUtil::utf8_to_pdf_doc(password, encoded)) { QTC::TC("qpdf", "QPDFJob password not encodable"); - throw std::runtime_error("supplied password cannot be encoded for" - " 40-bit or 128-bit encryption formats"); + throw std::runtime_error("supplied password cannot be encoded for 40-bit " + "or 128-bit encryption formats"); } password = encoded; } @@ -2687,31 +2627,27 @@ QPDFJob::maybeFixWritePassword(int R, std::string& password) if (QUtil::utf8_to_pdf_doc(password, encoded)) { QTC::TC("qpdf", "QPDFJob auto-encode password"); doIfVerbose([&](Pipeline& v, std::string const& prefix) { - v << prefix << ": automatically converting Unicode" - << " password to single-byte encoding as" - << " required for 40-bit or 128-bit" - << " encryption\n"; + v << prefix + << ": automatically converting Unicode password to single-byte " + "encoding as required for 40-bit or 128-bit encryption\n"; }); password = encoded; } else { QTC::TC("qpdf", "QPDFJob bytes fallback warning"); - *m->log->getError() << m->message_prefix << ": WARNING: " - << "supplied password looks like a Unicode" - << " password with characters not allowed in" - << " passwords for 40-bit and 128-bit " - "encryption;" - << " most readers will not be able to open this" - << " file with the supplied password." - << " (Use --password-mode=bytes to suppress " - "this" - << " warning and use the password anyway.)\n"; + *m->log->getError() + << m->message_prefix + << ": WARNING: supplied password looks like a Unicode password with " + "characters not allowed in passwords for 40-bit and 128-bit " + "encryption; most readers will not be able to open this file with " + "the supplied password. (Use --password-mode=bytes to suppress this " + "warning and use the password anyway.)\n"; } } else if ((R >= 5) && (!is_valid_utf8)) { QTC::TC("qpdf", "QPDFJob invalid utf-8 in auto"); - throw std::runtime_error("supplied password is not a valid Unicode password," - " which is required for 256-bit encryption; to" - " really use this password, rerun with the" - " --password-mode=bytes option"); + throw std::runtime_error( + "supplied password is not a valid Unicode password, which is required for " + "256-bit encryption; to really use this password, rerun with the " + "--password-mode=bytes option"); } } } @@ -2749,16 +2685,12 @@ QPDFJob::setEncryptionOptions(QPDF& pdf, QPDFWriter& w) if ((R < 4) || ((R == 4) && (!m->use_aes))) { if (!m->allow_weak_crypto) { QTC::TC("qpdf", "QPDFJob weak crypto error"); - *m->log->getError() << m->message_prefix - << ": refusing to write a file with RC4, a weak " - "cryptographic " - "algorithm\n" - << "Please use 256-bit keys for better security.\n" - << "Pass --allow-weak-crypto to enable writing insecure " - "files.\n" - << "See also " - "https://qpdf.readthedocs.io/en/stable/" - "weak-crypto.html\n"; + *m->log->getError() + << m->message_prefix + << ": refusing to write a file with RC4, a weak cryptographic algorithm\n" + "Please use 256-bit keys for better security.\n" + "Pass --allow-weak-crypto to enable writing insecure files.\n" + "See also https://qpdf.readthedocs.io/en/stable/weak-crypto.html\n"; throw std::runtime_error("refusing to write a file with weak crypto"); } } @@ -2996,8 +2928,8 @@ QPDFJob::doSplitPages(QPDF& pdf) qpdf_e_damaged_pdf, "", 0, - ("Exception caught while fixing copied" - " annotations. This may be a qpdf bug." + + ("Exception caught while fixing copied annotations. This may be a qpdf " + "bug." + std::string("Exception: ") + e.what())); } } @@ -3032,12 +2964,10 @@ QPDFJob::writeOutfile(QPDF& pdf) { std::shared_ptr temp_out; if (m->replace_input) { - // Append but don't prepend to the path to generate a - // temporary name. This saves us from having to split the path - // by directory and non-directory. + // Append but don't prepend to the path to generate a temporary name. This saves us from + // having to split the path by directory and non-directory. temp_out = QUtil::make_shared_cstr(std::string(m->infilename.get()) + ".~qpdf-temp#"); - // m->outfilename will be restored to 0 before temp_out - // goes out of scope. + // m->outfilename will be restored to 0 before temp_out goes out of scope. m->outfilename = temp_out; } else if (strcmp(m->outfilename.get(), "-") == 0) { m->outfilename = nullptr; @@ -3045,14 +2975,14 @@ QPDFJob::writeOutfile(QPDF& pdf) if (m->json_version) { writeJSON(pdf); } else { - // QPDFWriter must have block scope so the output file will be - // closed after write() finishes. + // QPDFWriter must have block scope so the output file will be closed after write() + // finishes. QPDFWriter w(pdf); if (m->outfilename) { w.setOutputFilename(m->outfilename.get()); } else { - // saveToStandardOutput has already been called, but - // calling it again is defensive and harmless. + // saveToStandardOutput has already been called, but calling it again is defensive and + // harmless. m->log->saveToStandardOutput(true); w.setOutputPipeline(m->log->getSave().get()); } @@ -3096,8 +3026,7 @@ QPDFJob::writeOutfile(QPDF& pdf) void QPDFJob::writeJSON(QPDF& pdf) { - // File pipeline must have block scope so it will be closed - // after write. + // File pipeline must have block scope so it will be closed after write. std::shared_ptr fc; std::shared_ptr fp; if (m->outfilename.get()) { diff --git a/libqpdf/QPDFObjectHandle.cc b/libqpdf/QPDFObjectHandle.cc index 7270b76d..fa5e52e8 100644 --- a/libqpdf/QPDFObjectHandle.cc +++ b/libqpdf/QPDFObjectHandle.cc @@ -51,8 +51,7 @@ QPDFObjectHandle::StreamDataProvider::StreamDataProvider(bool supports_retry) : QPDFObjectHandle::StreamDataProvider::~StreamDataProvider() { - // Must be explicit and not inline -- see QPDF_DLL_CLASS in - // README-maintainer + // Must be explicit and not inline -- see QPDF_DLL_CLASS in README-maintainer } void @@ -155,16 +154,14 @@ QPDFObjectHandle::TokenFilter::writeToken(QPDFTokenizer::Token const& token) void QPDFObjectHandle::ParserCallbacks::handleObject(QPDFObjectHandle) { - throw std::logic_error("You must override one of the" - " handleObject methods in ParserCallbacks"); + throw std::logic_error("You must override one of the handleObject methods in ParserCallbacks"); } void QPDFObjectHandle::ParserCallbacks::handleObject(QPDFObjectHandle oh, size_t, size_t) { - // This version of handleObject was added in qpdf 9. If the - // developer did not override it, fall back to the older - // interface. + // This version of handleObject was added in qpdf 9. If the developer did not override it, fall + // back to the older interface. handleObject(oh); } @@ -592,8 +589,7 @@ QPDFObjectHandle::getUIntValueAsUInt() result = 0; } else if (v > UINT_MAX) { QTC::TC("qpdf", "QPDFObjectHandle uint returning UINT_MAX"); - warnIfPossible("requested value of unsigned integer is too big;" - " returning UINT_MAX"); + warnIfPossible("requested value of unsigned integer is too big; returning UINT_MAX"); result = UINT_MAX; } else { result = static_cast(v); @@ -1092,11 +1088,9 @@ QPDFObjectHandle::mergeResources( QPDFObjectHandle this_val = getKey(rtype); if (this_val.isDictionary() && other_val.isDictionary()) { if (this_val.isIndirect()) { - // Do this even if there are no keys. Various - // places in the code call mergeResources with - // resource dictionaries that contain empty - // subdictionaries just to get this shallow copy - // functionality. + // Do this even if there are no keys. Various places in the code call + // mergeResources with resource dictionaries that contain empty subdictionaries + // just to get this shallow copy functionality. QTC::TC("qpdf", "QPDFObjectHandle replace with copy"); this_val = replaceKeyAndGetNew(rtype, this_val.shallowCopy()); } @@ -1476,8 +1470,7 @@ QPDFObjectHandle::arrayOrStreamToStreamArray( "", description, 0, - " object is supposed to be a stream or an" - " array of streams but is neither")); + " object is supposed to be a stream or an array of streams but is neither")); } bool first = true; @@ -1526,8 +1519,8 @@ void QPDFObjectHandle::rotatePage(int angle, bool relative) { if ((angle % 90) != 0) { - throw std::runtime_error("QPDF::rotatePage called with an" - " angle that is not a multiple of 90"); + throw std::runtime_error( + "QPDF::rotatePage called with an angle that is not a multiple of 90"); } int new_angle = angle; if (relative) { @@ -1551,8 +1544,7 @@ QPDFObjectHandle::rotatePage(int angle, bool relative) new_angle += old_angle; } new_angle = (new_angle + 360) % 360; - // Make this explicit even with new_angle == 0 since /Rotate can - // be inherited. + // Make this explicit even with new_angle == 0 since /Rotate can be inherited. replaceKey("/Rotate", QPDFObjectHandle::newInteger(new_angle)); } @@ -1564,15 +1556,14 @@ QPDFObjectHandle::coalesceContentStreams() QTC::TC("qpdf", "QPDFObjectHandle coalesce called on stream"); return; } else if (!contents.isArray()) { - // /Contents is optional for pages, and some very damaged - // files may have pages that are invalid in other ways. + // /Contents is optional for pages, and some very damaged files may have pages that are + // invalid in other ways. return; } - // Should not be possible for a page object to not have an - // owning PDF unless it was manually constructed in some - // incorrect way. However, it can happen in a PDF file whose - // page structure is direct, which is against spec but still - // possible to hand construct, as in fuzz issue 27393. + // Should not be possible for a page object to not have an owning PDF unless it was manually + // constructed in some incorrect way. However, it can happen in a PDF file whose page structure + // is direct, which is against spec but still possible to hand construct, as in fuzz issue + // 27393. QPDF& qpdf = getQPDF("coalesceContentStreams called on object with no associated PDF file"); QPDFObjectHandle new_contents = newStream(&qpdf); @@ -1808,8 +1799,8 @@ QPDFObjectHandle::parseContentStream_data( callbacks->handleObject(obj, QIntC::to_size(offset), length); if (obj.isOperator() && (obj.getOperatorValue() == "ID")) { - // Discard next character; it is the space after ID that - // terminated the token. Read until end of inline image. + // Discard next character; it is the space after ID that terminated the token. Read + // until end of inline image. char ch; input->read(&ch, 1); tokenizer.expectInlineImage(input); @@ -2052,8 +2043,8 @@ QPDFObjectHandle::newReserved(QPDF* qpdf) void QPDFObjectHandle::setObjectDescription(QPDF* owning_qpdf, std::string const& object_description) { - // This is called during parsing on newly created direct objects, - // so we can't call dereference() here. + // This is called during parsing on newly created direct objects, so we can't call dereference() + // here. if (isInitialized() && obj.get()) { auto descr = std::make_shared(object_description); obj->setDescription(owning_qpdf, descr); @@ -2070,8 +2061,7 @@ QPDFObjectHandle QPDFObjectHandle::shallowCopy() { if (!dereference()) { - throw std::logic_error("operation attempted on uninitialized " - "QPDFObjectHandle"); + throw std::logic_error("operation attempted on uninitialized QPDFObjectHandle"); } return QPDFObjectHandle(obj->copy()); } @@ -2080,8 +2070,7 @@ QPDFObjectHandle QPDFObjectHandle::unsafeShallowCopy() { if (!dereference()) { - throw std::logic_error("operation attempted on uninitialized " - "QPDFObjectHandle"); + throw std::logic_error("operation attempted on uninitialized QPDFObjectHandle"); } return QPDFObjectHandle(obj->copy(true)); } @@ -2094,8 +2083,7 @@ QPDFObjectHandle::makeDirect(QPDFObjGen::set& visited, bool stop_at_streams) auto cur_og = getObjGen(); if (!visited.add(cur_og)) { QTC::TC("qpdf", "QPDFObjectHandle makeDirect loop"); - throw std::runtime_error("loop detected while converting object from " - "indirect to direct"); + throw std::runtime_error("loop detected while converting object from indirect to direct"); } if (isBool() || isInteger() || isName() || isNull() || isReal() || isString()) { @@ -2123,11 +2111,10 @@ QPDFObjectHandle::makeDirect(QPDFObjGen::set& visited, bool stop_at_streams) throw std::runtime_error("attempt to make a stream into a direct object"); } } else if (isReserved()) { - throw std::logic_error("QPDFObjectHandle: attempting to make a" - " reserved object handle direct"); + throw std::logic_error( + "QPDFObjectHandle: attempting to make a reserved object handle direct"); } else { - throw std::logic_error("QPDFObjectHandle::makeDirectInternal: " - "unknown object type"); + throw std::logic_error("QPDFObjectHandle::makeDirectInternal: unknown object type"); } visited.erase(cur_og); @@ -2162,8 +2149,7 @@ void QPDFObjectHandle::assertInitialized() const { if (!isInitialized()) { - throw std::logic_error("operation attempted on uninitialized " - "QPDFObjectHandle"); + throw std::logic_error("operation attempted on uninitialized QPDFObjectHandle"); } } @@ -2172,8 +2158,8 @@ QPDFObjectHandle::typeWarning(char const* expected_type, std::string const& warn { QPDF* context = nullptr; std::string description; - // Type checks above guarantee that the object has been dereferenced. - // Nevertheless, dereference throws exceptions in the test suite + // Type checks above guarantee that the object has been dereferenced. Nevertheless, dereference + // throws exceptions in the test suite if (!dereference()) { throw std::logic_error("attempted to dereference an uninitialized QPDFObjectHandle"); } @@ -2376,8 +2362,8 @@ QPDFObjectHandle::checkOwnership(QPDFObjectHandle const& item) const auto item_qpdf = item.getOwningQPDF(); if ((qpdf != nullptr) && (item_qpdf != nullptr) && (qpdf != item_qpdf)) { QTC::TC("qpdf", "QPDFObjectHandle check ownership"); - throw std::logic_error("Attempting to add an object from a different QPDF." - " Use QPDF::copyForeignObject to add objects from another file."); + throw std::logic_error("Attempting to add an object from a different QPDF. Use " + "QPDF::copyForeignObject to add objects from another file."); } } @@ -2402,9 +2388,8 @@ QPDFObjectHandle::dereference() void QPDFObjectHandle::warn(QPDF* qpdf, QPDFExc const& e) { - // If parsing on behalf of a QPDF object and want to give a - // warning, we can warn through the object. If parsing for some - // other reason, such as an explicit creation of an object from a + // If parsing on behalf of a QPDF object and want to give a warning, we can warn through the + // object. If parsing for some other reason, such as an explicit creation of an object from a // string, then just throw the exception. if (qpdf) { qpdf->warn(e); @@ -2596,7 +2581,8 @@ QPDFObjectHandle::getQPDF(std::string const& error_msg) const { auto result = isInitialized() ? this->obj->getQPDF() : nullptr; if (result == nullptr) { - throw std::runtime_error(error_msg == "" ? "attempt to use a null qpdf object" : error_msg); + throw std::runtime_error( + error_msg.empty() ? "attempt to use a null qpdf object" : error_msg); } return *result; } diff --git a/libqpdf/QPDFPageObjectHelper.cc b/libqpdf/QPDFPageObjectHelper.cc index a34ed28f..608254e4 100644 --- a/libqpdf/QPDFPageObjectHelper.cc +++ b/libqpdf/QPDFPageObjectHelper.cc @@ -110,10 +110,8 @@ InlineImageTracker::convertIIDict(QPDFObjectHandle odict) } else if (name == "/I") { name = "/Indexed"; } else { - // This is a key in the page's /Resources -> - // /ColorSpace dictionary. We need to look it up - // and use its value as the color space for the - // image. + // This is a key in the page's /Resources -> /ColorSpace dictionary. We need to + // look it up and use its value as the color space for the image. QPDFObjectHandle colorspace = resources.getKey("/ColorSpace"); if (colorspace.isDictionary() && colorspace.hasKey(name)) { QTC::TC("qpdf", "QPDFPageObjectHelper colorspace lookup"); @@ -407,8 +405,8 @@ QPDFPageObjectHelper::externalizeInlineImages(size_t min_size, bool shallow) { if (shallow) { QPDFObjectHandle resources = getAttribute("/Resources", true); - // Calling mergeResources also ensures that /XObject becomes - // direct and is not shared with other pages. + // Calling mergeResources also ensures that /XObject becomes direct and is not shared with + // other pages. resources.mergeResources("<< /XObject << >> >>"_qpdf); InlineImageTracker iit(this->oh.getOwningQPDF(), min_size, resources); Pl_Buffer b("new page content"); @@ -573,11 +571,10 @@ QPDFPageObjectHelper::removeUnreferencedResourcesHelper( return false; } - // We will walk through /Font and /XObject dictionaries, removing - // any resources that are not referenced. We must make copies of - // resource dictionaries down into the dictionaries are mutating - // to prevent mutating one dictionary from having the side effect - // of mutating the one it was copied from. + // We will walk through /Font and /XObject dictionaries, removing any resources that are not + // referenced. We must make copies of resource dictionaries down into the dictionaries are + // mutating to prevent mutating one dictionary from having the side effect of mutating the one + // it was copied from. QPDFObjectHandle resources = ph.getAttribute("/Resources", true); std::vector rdicts; std::set known_names; @@ -605,33 +602,25 @@ QPDFPageObjectHelper::removeUnreferencedResourcesHelper( } } } - // Older versions of the PDF spec allowed form XObjects to omit - // their resources dictionaries, in which case names were resolved - // from the containing page. This behavior seems to be widely - // supported by viewers. If a form XObjects has a resources - // dictionary and has some unresolved names, some viewers fail to - // resolve them, and others allow them to be inherited from the - // page or from another form XObjects that contains them. Since - // this behavior is inconsistent across viewers, we consider an - // unresolved name when a resources dictionary is present to be - // reason not to remove unreferenced resources. An unresolved name - // in the absence of a resource dictionary is not considered a - // problem. For form XObjects, we just accumulate a list of - // unresolved names, and for page objects, we avoid removing any - // such names found in nested form XObjects. + // Older versions of the PDF spec allowed form XObjects to omit their resources dictionaries, in + // which case names were resolved from the containing page. This behavior seems to be widely + // supported by viewers. If a form XObjects has a resources dictionary and has some unresolved + // names, some viewers fail to resolve them, and others allow them to be inherited from the page + // or from another form XObjects that contains them. Since this behavior is inconsistent across + // viewers, we consider an unresolved name when a resources dictionary is present to be reason + // not to remove unreferenced resources. An unresolved name in the absence of a resource + // dictionary is not considered a problem. For form XObjects, we just accumulate a list of + // unresolved names, and for page objects, we avoid removing any such names found in nested form + // XObjects. if ((!local_unresolved.empty()) && resources.isDictionary()) { - // It's not worth issuing a warning for this case. From qpdf - // 10.3, we are hopefully only looking at names that are - // referencing fonts and XObjects, but until we're certain - // that we know the meaning of every name in a content stream, - // we don't want to give warnings that might be false - // positives. Also, this can happen in legitimate cases with - // older PDFs, and there's nothing to be done about it, so - // there's no good reason to issue a warning. The only sad - // thing is that it was a false positive that alerted me to a - // logic error in the code, and any future such errors would - // now be hidden. + // It's not worth issuing a warning for this case. From qpdf 10.3, we are hopefully only + // looking at names that are referencing fonts and XObjects, but until we're certain that we + // know the meaning of every name in a content stream, we don't want to give warnings that + // might be false positives. Also, this can happen in legitimate cases with older PDFs, and + // there's nothing to be done about it, so there's no good reason to issue a warning. The + // only sad thing is that it was a false positive that alerted me to a logic error in the + // code, and any future such errors would now be hidden. QTC::TC("qpdf", "QPDFPageObjectHelper unresolved names"); return false; } @@ -639,8 +628,7 @@ QPDFPageObjectHelper::removeUnreferencedResourcesHelper( for (auto& dict: rdicts) { for (auto const& key: dict.getKeys()) { if (is_page && unresolved.count(key)) { - // This name is referenced by some nested form - // xobject, so don't remove it. + // This name is referenced by some nested form xobject, so don't remove it. QTC::TC("qpdf", "QPDFPageObjectHelper resolving unresolved"); } else if (!rf.getNames().count(key)) { dict.removeKey(key); @@ -653,8 +641,7 @@ QPDFPageObjectHelper::removeUnreferencedResourcesHelper( void QPDFPageObjectHelper::removeUnreferencedResources() { - // Accumulate a list of unresolved names across all nested form - // XObjects. + // Accumulate a list of unresolved names across all nested form XObjects. std::set unresolved; bool any_failures = false; forEachFormXObject( @@ -724,10 +711,9 @@ QPDFPageObjectHelper::getMatrixForTransformations(bool invert) QPDFObjectHandle QPDFPageObjectHelper::getFormXObjectForPage(bool handle_transformations) { - auto result = this->oh - .getQPDF("QPDFPageObjectHelper::getFormXObjectForPage " - "called with a direct object") - .newStream(); + auto result = + this->oh.getQPDF("QPDFPageObjectHelper::getFormXObjectForPage called with a direct object") + .newStream(); QPDFObjectHandle newdict = result.getDict(); newdict.replaceKey("/Type", QPDFObjectHandle::newName("/XObject")); newdict.replaceKey("/Subtype", QPDFObjectHandle::newName("/Form")); @@ -759,18 +745,15 @@ QPDFPageObjectHelper::getMatrixForFormXObjectPlacement( bool allow_shrink, bool allow_expand) { - // Calculate the transformation matrix that will place the given - // form XObject fully inside the given rectangle, center and - // shrinking or expanding as needed if requested. - - // When rendering a form XObject, the transformation in the - // graphics state (cm) is applied first (of course -- when it is - // applied, the PDF interpreter doesn't even know we're going to - // be drawing a form XObject yet), and then the object's matrix - // (M) is applied. The resulting matrix, when applied to the form - // XObject's bounding box, will generate a new rectangle. We want - // to create a transformation matrix that make the form XObject's - // bounding box land in exactly the right spot. + // Calculate the transformation matrix that will place the given form XObject fully inside the + // given rectangle, center and shrinking or expanding as needed if requested. + + // When rendering a form XObject, the transformation in the graphics state (cm) is applied first + // (of course -- when it is applied, the PDF interpreter doesn't even know we're going to be + // drawing a form XObject yet), and then the object's matrix (M) is applied. The resulting + // matrix, when applied to the form XObject's bounding box, will generate a new rectangle. We + // want to create a transformation matrix that make the form XObject's bounding box land in + // exactly the right spot. QPDFObjectHandle fdict = fo.getDict(); QPDFObjectHandle bbox_obj = fdict.getKey("/BBox"); @@ -782,37 +765,32 @@ QPDFPageObjectHelper::getMatrixForFormXObjectPlacement( QPDFMatrix tmatrix; // "to" matrix QPDFMatrix fmatrix; // "from" matrix if (invert_transformations) { - // tmatrix inverts scaling and rotation of the destination - // page. Applying this matrix allows the overlaid form - // XObject's to be absolute rather than relative to properties - // of the destination page. tmatrix is part of the computed - // transformation matrix. + // tmatrix inverts scaling and rotation of the destination page. Applying this matrix allows + // the overlaid form XObject's to be absolute rather than relative to properties of the + // destination page. tmatrix is part of the computed transformation matrix. tmatrix = QPDFMatrix(getMatrixForTransformations(true)); wmatrix.concat(tmatrix); } if (fdict.getKey("/Matrix").isMatrix()) { - // fmatrix is the transformation matrix that is applied to the - // form XObject itself. We need this for calculations, but we - // don't explicitly use it in the final result because the PDF + // fmatrix is the transformation matrix that is applied to the form XObject itself. We need + // this for calculations, but we don't explicitly use it in the final result because the PDF // rendering system automatically applies this last before // drawing the form XObject. fmatrix = QPDFMatrix(fdict.getKey("/Matrix").getArrayAsMatrix()); wmatrix.concat(fmatrix); } - // The current wmatrix handles transformation from the form - // xobject and, if requested, the destination page. Next, we have - // to adjust this for scale and position. + // The current wmatrix handles transformation from the form xobject and, if requested, the + // destination page. Next, we have to adjust this for scale and position. - // Step 1: figure out what scale factor we need to make the form - // XObject's bounding box fit within the destination rectangle. + // Step 1: figure out what scale factor we need to make the form XObject's bounding box fit + // within the destination rectangle. // Transform bounding box QPDFObjectHandle::Rectangle bbox = bbox_obj.getArrayAsRectangle(); QPDFObjectHandle::Rectangle T = wmatrix.transformRectangle(bbox); - // Calculate a scale factor, if needed. Shrink or expand if needed - // and allowed. + // Calculate a scale factor, if needed. Shrink or expand if needed and allowed. if ((T.urx == T.llx) || (T.ury == T.lly)) { // avoid division by zero return QPDFMatrix(); @@ -834,8 +812,8 @@ QPDFPageObjectHelper::getMatrixForFormXObjectPlacement( } } - // Step 2: figure out what translation is required to get the - // rectangle to the right spot: centered within the destination. + // Step 2: figure out what translation is required to get the rectangle to the right spot: + // centered within the destination. wmatrix = QPDFMatrix(); wmatrix.scale(scale, scale); wmatrix.concat(tmatrix); @@ -849,9 +827,8 @@ QPDFPageObjectHelper::getMatrixForFormXObjectPlacement( double tx = r_cx - t_cx; double ty = r_cy - t_cy; - // Now we can calculate the final matrix. The final matrix does - // not include fmatrix because that is applied automatically by - // the PDF interpreter. + // Now we can calculate the final matrix. The final matrix does not include fmatrix because that + // is applied automatically by the PDF interpreter. QPDFMatrix cm; cm.translate(tx, ty); cm.scale(scale, scale); @@ -921,18 +898,15 @@ QPDFPageObjectHelper::flattenRotation(QPDFAcroFormDocumentHelper* afdh) auto rect = box.getArrayAsRectangle(); decltype(rect) new_rect; - // How far are the edges of our rectangle from the edges - // of the media box? + // How far are the edges of our rectangle from the edges of the media box? auto left_x = rect.llx - media_rect.llx; auto right_x = media_rect.urx - rect.urx; auto bottom_y = rect.lly - media_rect.lly; auto top_y = media_rect.ury - rect.ury; - // Rotating the page 180 degrees does not change - // /MediaBox. Rotating 90 or 270 degrees reverses llx and - // lly and also reverse urx and ury. For all the other - // boxes, we want the corners to be the correct distance - // away from the corners of the mediabox. + // Rotating the page 180 degrees does not change /MediaBox. Rotating 90 or 270 degrees + // reverses llx and lly and also reverse urx and ury. For all the other boxes, we want the + // corners to be the correct distance away from the corners of the mediabox. switch (rotate) { case 90: new_rect.llx = media_rect.lly + bottom_y; @@ -963,9 +937,8 @@ QPDFPageObjectHelper::flattenRotation(QPDFAcroFormDocumentHelper* afdh) this->oh.replaceKey(boxkey, QPDFObjectHandle::newFromRectangle(new_rect)); } - // When we rotate the page, pivot about the point 0, 0 and then - // translate so the page is visible with the origin point being - // the same offset from the lower left corner of the media box. + // When we rotate the page, pivot about the point 0, 0 and then translate so the page is visible + // with the origin point being the same offset from the lower left corner of the media box. // These calculations have been verified empirically with various // PDF readers. QPDFMatrix cm(0, 0, 0, 0, 0, 0); diff --git a/libqpdf/QPDFParser.cc b/libqpdf/QPDFParser.cc index 5d695897..48227e55 100644 --- a/libqpdf/QPDFParser.cc +++ b/libqpdf/QPDFParser.cc @@ -41,12 +41,10 @@ namespace QPDFObjectHandle QPDFParser::parse(bool& empty, bool content_stream) { - // This method must take care not to resolve any objects. Don't - // check the type of any object without first ensuring that it is - // a direct object. Otherwise, doing so may have the side effect - // of reading the object and changing the file pointer. If you do - // this, it will cause a logic error to be thrown from - // QPDF::inParse(). + // This method must take care not to resolve any objects. Don't check the type of any object + // without first ensuring that it is a direct object. Otherwise, doing so may have the side + // effect of reading the object and changing the file pointer. If you do this, it will cause a + // logic error to be thrown from QPDF::inParse(). const static std::shared_ptr null_oh = QPDF_Null::create(); QPDF::ParseGuard pg(context); @@ -193,18 +191,16 @@ QPDFParser::parse(bool& empty, bool content_stream) !olist.at(size - 2)->getObjGen().isIndirect()) { if (context == nullptr) { QTC::TC("qpdf", "QPDFParser indirect without context"); - throw std::logic_error("QPDFObjectHandle::parse called without context" - " on an object with indirect references"); + throw std::logic_error("QPDFObjectHandle::parse called without context on " + "an object with indirect references"); } auto ref_og = QPDFObjGen( QPDFObjectHandle(olist.at(size - 2)).getIntValueAsInt(), QPDFObjectHandle(olist.back()).getIntValueAsInt()); if (ref_og.isIndirect()) { - // This action has the desirable side effect - // of causing dangling references (references - // to indirect objects that don't appear in - // the PDF) in any parsed object to appear in - // the object cache. + // This action has the desirable side effect of causing dangling references + // (references to indirect objects that don't appear in the PDF) in any + // parsed object to appear in the object cache. object = context->getObject(ref_og).obj; indirect_ref = true; } else { @@ -214,16 +210,14 @@ QPDFParser::parse(bool& empty, bool content_stream) olist.pop_back(); olist.pop_back(); } else if ((value == "endobj") && (state == st_top)) { - // We just saw endobj without having read - // anything. Treat this as a null and do not move - // the input source's offset. + // We just saw endobj without having read anything. Treat this as a null and do + // not move the input source's offset. is_null = true; input->seek(input->getLastOffset(), SEEK_SET); empty = true; } else { QTC::TC("qpdf", "QPDFParser treat word as string"); - warn("unknown token while reading object;" - " treating as string"); + warn("unknown token while reading object; treating as string"); bad = true; object = QPDF_String::create(value); } @@ -250,8 +244,7 @@ QPDFParser::parse(bool& empty, bool content_stream) break; default: - warn("treating unknown token type as null while " - "reading object"); + warn("treating unknown token type as null while reading object"); bad = true; is_null = true; break; @@ -259,8 +252,7 @@ QPDFParser::parse(bool& empty, bool content_stream) if (object == nullptr && !is_null && (!((state == st_start) || (state == st_stop) || (state == st_eof)))) { - throw std::logic_error("QPDFObjectHandle::parseInternal: " - "unexpected uninitialized object"); + throw std::logic_error("QPDFParser:parseInternal: unexpected uninitialized object"); is_null = true; } @@ -274,8 +266,8 @@ QPDFParser::parse(bool& empty, bool content_stream) } } if (bad_count > 5) { - // We had too many consecutive errors without enough - // intervening successful objects. Give up. + // We had too many consecutive errors without enough intervening successful objects. + // Give up. warn("too many errors; giving up on reading object"); state = st_top; is_null = true; @@ -287,8 +279,7 @@ QPDFParser::parse(bool& empty, bool content_stream) warn("parse error while reading object"); } done = true; - // In content stream mode, leave object uninitialized to - // indicate EOF + // In content stream mode, leave object uninitialized to indicate EOF if (!content_stream) { is_null = true; } @@ -298,8 +289,7 @@ QPDFParser::parse(bool& empty, bool content_stream) case st_array: if (is_null) { object = null_oh; - // No need to set description for direct nulls - they probably - // will become implicit. + // No need to set description for direct nulls - they probably will become implicit. } else if (!indirect_ref) { setDescription(object, input->getLastOffset()); } @@ -316,23 +306,22 @@ QPDFParser::parse(bool& empty, bool content_stream) case st_stop: if ((state_stack.size() < 2) || (stack.size() < 2)) { - throw std::logic_error("QPDFObjectHandle::parseInternal: st_stop encountered" - " with insufficient elements in stack"); + throw std::logic_error("QPDFParser::parseInternal: st_stop encountered with " + "insufficient elements in stack"); } parser_state_e old_state = state_stack.back(); state_stack.pop_back(); if (old_state == st_array) { object = QPDF_Array::create(std::move(olist), frame.null_count > 100); setDescription(object, offset - 1); - // The `offset` points to the next of "[". Set the rewind - // offset to point to the beginning of "[". This has been - // explicitly tested with whitespace surrounding the array start - // delimiter. getLastOffset points to the array end token and - // therefore can't be used here. + // The `offset` points to the next of "[". Set the rewind offset to point to the + // beginning of "[". This has been explicitly tested with whitespace surrounding the + // array start delimiter. getLastOffset points to the array end token and therefore + // can't be used here. set_offset = true; } else if (old_state == st_dictionary) { - // Convert list to map. Alternating elements are keys. Attempt - // to recover more or less gracefully from invalid dictionaries. + // Convert list to map. Alternating elements are keys. Attempt to recover more or + // less gracefully from invalid dictionaries. std::set names; for (auto& obj: olist) { if (obj) { @@ -358,8 +347,7 @@ QPDFParser::parse(bool& empty, bool content_stream) } warn( offset, - "expected dictionary key but found" - " non-name object; inserting key " + + "expected dictionary key but found non-name object; inserting key " + key); } if (dict.count(key) > 0) { @@ -367,8 +355,7 @@ QPDFParser::parse(bool& empty, bool content_stream) warn( offset, "dictionary has duplicated key " + key + - "; last occurrence overrides earlier " - "ones"); + "; last occurrence overrides earlier ones"); } // Calculate value. @@ -380,8 +367,7 @@ QPDFParser::parse(bool& empty, bool content_stream) QTC::TC("qpdf", "QPDFParser no val for last key"); warn( offset, - "dictionary ended prematurely; " - "using null as value for last key"); + "dictionary ended prematurely; using null as value for last key"); val = QPDF_Null::create(); } @@ -395,11 +381,10 @@ QPDFParser::parse(bool& empty, bool content_stream) } object = QPDF_Dictionary::create(std::move(dict)); setDescription(object, offset - 2); - // The `offset` points to the next of "<<". Set the rewind - // offset to point to the beginning of "<<". This has been - // explicitly tested with whitespace surrounding the dictionary - // start delimiter. getLastOffset points to the dictionary end - // token and therefore can't be used here. + // The `offset` points to the next of "<<". Set the rewind offset to point to the + // beginning of "<<". This has been explicitly tested with whitespace surrounding + // the dictionary start delimiter. getLastOffset points to the dictionary end token + // and therefore can't be used here. set_offset = true; } stack.pop_back(); @@ -431,9 +416,8 @@ QPDFParser::setDescription(std::shared_ptr& obj, qpdf_offset_t parse void QPDFParser::warn(QPDFExc const& e) const { - // If parsing on behalf of a QPDF object and want to give a - // warning, we can warn through the object. If parsing for some - // other reason, such as an explicit creation of an object from a + // If parsing on behalf of a QPDF object and want to give a warning, we can warn through the + // object. If parsing for some other reason, such as an explicit creation of an object from a // string, then just throw the exception. if (context) { context->warn(e); diff --git a/libqpdf/QPDFTokenizer.cc b/libqpdf/QPDFTokenizer.cc index da02a0fe..d98af8a9 100644 --- a/libqpdf/QPDFTokenizer.cc +++ b/libqpdf/QPDFTokenizer.cc @@ -1,8 +1,7 @@ #include -// DO NOT USE ctype -- it is locale dependent for some things, and -// it's not worth the risk of including it in case it may accidentally -// be used. +// DO NOT USE ctype -- it is locale dependent for some things, and it's not worth the risk of +// including it in case it may accidentally be used. #include #include @@ -45,8 +44,8 @@ namespace bool QPDFWordTokenFinder::check() { - // Find a word token matching the given string, preceded by a - // delimiter, and followed by a delimiter or EOF. + // Find a word token matching the given string, preceded by a delimiter, and followed by a + // delimiter or EOF. QPDFTokenizer tokenizer; QPDFTokenizer::Token t = tokenizer.readToken(is, "finder", true); qpdf_offset_t pos = is->tell(); @@ -68,8 +67,7 @@ QPDFWordTokenFinder::check() return false; } if (token_start == 0) { - // Can't actually happen...we never start the search at the - // beginning of the input. + // Can't actually happen...we never start the search at the beginning of the input. return false; } return true; @@ -147,9 +145,9 @@ QPDFTokenizer::presentCharacter(char ch) void QPDFTokenizer::handleCharacter(char ch) { - // State machine is implemented such that the final character may not be - // handled. This happens whenever you have to use a character from the - // next token to detect the end of the current token. + // State machine is implemented such that the final character may not be handled. This happens + // whenever you have to use a character from the next token to detect the end of the current + // token. switch (this->state) { case st_top: @@ -248,15 +246,14 @@ QPDFTokenizer::handleCharacter(char ch) void QPDFTokenizer::inTokenReady(char ch) { - throw std::logic_error("INTERNAL ERROR: QPDF tokenizer presented character " - "while token is waiting"); + throw std::logic_error( + "INTERNAL ERROR: QPDF tokenizer presented character while token is waiting"); } void QPDFTokenizer::inBeforeToken(char ch) { - // Note: we specifically do not use ctype here. It is - // locale-dependent. + // Note: we specifically do not use ctype here. It is locale-dependent. if (isSpace(ch)) { this->before_token = !this->include_ignorable; this->in_token = this->include_ignorable; @@ -421,11 +418,9 @@ void QPDFTokenizer::inName(char ch) { if (isDelimiter(ch)) { - // A C-locale whitespace character or delimiter terminates - // token. It is important to unread the whitespace - // character even though it is ignored since it may be the - // newline after a stream keyword. Removing it here could - // make the stream-reading code break on some files, + // A C-locale whitespace character or delimiter terminates token. It is important to unread + // the whitespace character even though it is ignored since it may be the newline after a + // stream keyword. Removing it here could make the stream-reading code break on some files, // though not on any files in the test suite as of this // writing. @@ -452,8 +447,7 @@ QPDFTokenizer::inNameHex1(char ch) } else { QTC::TC("qpdf", "QPDFTokenizer bad name 1"); this->error_message = "name with stray # will not work with PDF >= 1.2"; - // Use null to encode a bad # -- this is reversed - // in QPDF_Name::normalizeName. + // Use null to encode a bad # -- this is reversed in QPDF_Name::normalizeName. this->val += '\0'; this->state = st_name; inName(ch); @@ -468,8 +462,7 @@ QPDFTokenizer::inNameHex2(char ch) } else { QTC::TC("qpdf", "QPDFTokenizer bad name 2"); this->error_message = "name with stray # will not work with PDF >= 1.2"; - // Use null to encode a bad # -- this is reversed - // in QPDF_Name::normalizeName. + // Use null to encode a bad # -- this is reversed in QPDF_Name::normalizeName. this->val += '\0'; this->val += this->hex_char; this->state = st_name; @@ -636,13 +629,10 @@ void QPDFTokenizer::inLiteral(char ch) { if (isDelimiter(ch)) { - // A C-locale whitespace character or delimiter terminates - // token. It is important to unread the whitespace - // character even though it is ignored since it may be the - // newline after a stream keyword. Removing it here could - // make the stream-reading code break on some files, - // though not on any files in the test suite as of this - // writing. + // A C-locale whitespace character or delimiter terminates token. It is important to unread + // the whitespace character even though it is ignored since it may be the newline after a + // stream keyword. Removing it here could make the stream-reading code break on some files, + // though not on any files in the test suite as of this writing. this->in_token = false; this->char_to_unread = ch; @@ -707,8 +697,7 @@ QPDFTokenizer::inCharCode(char ch) if (++(this->digit_count) < 3) { return; } - // We've accumulated \ddd. PDF Spec says to ignore - // high-order overflow. + // We've accumulated \ddd. PDF Spec says to ignore high-order overflow. } this->val += char(this->char_code % 256); this->state = st_in_string; @@ -739,8 +728,7 @@ QPDFTokenizer::presentEOF() case st_decimal: case st_literal: QTC::TC("qpdf", "QPDFTokenizer EOF reading appendable token"); - // Push any delimiter to the state machine to finish off the final - // token. + // Push any delimiter to the state machine to finish off the final token. presentCharacter('\f'); this->in_token = true; break; @@ -794,14 +782,12 @@ QPDFTokenizer::findEI(std::shared_ptr input) qpdf_offset_t last_offset = input->getLastOffset(); qpdf_offset_t pos = input->tell(); - // Use QPDFWordTokenFinder to find EI surrounded by delimiters. - // Then read the next several tokens or up to EOF. If we find any - // suspicious-looking or tokens, this is probably still part of - // the image data, so keep looking for EI. Stop at the first EI - // that passes. If we get to the end without finding one, return - // the last EI we found. Store the number of bytes expected in the - // inline image including the EI and use that to break out of - // inline image, falling back to the old method if needed. + // Use QPDFWordTokenFinder to find EI surrounded by delimiters. Then read the next several + // tokens or up to EOF. If we find any suspicious-looking or tokens, this is probably still part + // of the image data, so keep looking for EI. Stop at the first EI that passes. If we get to the + // end without finding one, return the last EI we found. Store the number of bytes expected in + // the inline image including the EI and use that to break out of inline image, falling back to + // the old method if needed. bool okay = false; bool first_try = true; @@ -814,13 +800,11 @@ QPDFTokenizer::findEI(std::shared_ptr input) QPDFTokenizer check; bool found_bad = false; - // Look at the next 10 tokens or up to EOF. The next inline - // image's image data would look like bad tokens, but there - // will always be at least 10 tokens between one inline - // image's EI and the next valid one's ID since width, height, - // bits per pixel, and color space are all required as well as - // a BI and ID. If we get 10 good tokens in a row or hit EOF, - // we can be pretty sure we've found the actual EI. + // Look at the next 10 tokens or up to EOF. The next inline image's image data would look + // like bad tokens, but there will always be at least 10 tokens between one inline image's + // EI and the next valid one's ID since width, height, bits per pixel, and color space are + // all required as well as a BI and ID. If we get 10 good tokens in a row or hit EOF, we can + // be pretty sure we've found the actual EI. for (int i = 0; i < 10; ++i) { QPDFTokenizer::Token t = check.readToken(input, "checker", true); token_type_e type = t.getType(); @@ -829,27 +813,22 @@ QPDFTokenizer::findEI(std::shared_ptr input) } else if (type == tt_bad) { found_bad = true; } else if (t.isWord()) { - // The qpdf tokenizer lumps alphabetic and otherwise - // uncategorized characters into "words". We recognize - // strings of alphabetic characters as potential valid - // operators for purposes of telling whether we're in - // valid content or not. It's not perfect, but it - // should work more reliably than what we used to do, - // which was already good enough for the vast majority - // of files. + // The qpdf tokenizer lumps alphabetic and otherwise uncategorized characters into + // "words". We recognize strings of alphabetic characters as potential valid + // operators for purposes of telling whether we're in valid content or not. It's not + // perfect, but it should work more reliably than what we used to do, which was + // already good enough for the vast majority of files. bool found_alpha = false; bool found_non_printable = false; bool found_other = false; for (char ch: t.getValue()) { if (((ch >= 'a') && (ch <= 'z')) || ((ch >= 'A') && (ch <= 'Z')) || (ch == '*')) { - // Treat '*' as alpha since there are valid - // PDF operators that contain * along with - // alphabetic characters. + // Treat '*' as alpha since there are valid PDF operators that contain * + // along with alphabetic characters. found_alpha = true; } else if ((static_cast(ch) < 32) && (!isSpace(ch))) { - // Compare ch as a signed char so characters - // outside of 7-bit will be < 0. + // Compare ch as a signed char so characters outside of 7-bit will be < 0. found_non_printable = true; break; } else { @@ -903,9 +882,9 @@ QPDFTokenizer::betweenTokens() QPDFTokenizer::Token QPDFTokenizer::readToken( - std::shared_ptr input, std::string const& context, bool allow_bad, size_t max_len) + InputSource& input, std::string const& context, bool allow_bad, size_t max_len) { - nextToken(*input, context, max_len); + nextToken(input, context, max_len); Token token; bool unread_char; @@ -918,15 +897,22 @@ QPDFTokenizer::readToken( } else { throw QPDFExc( qpdf_e_damaged_pdf, - input->getName(), + input.getName(), context, - input->getLastOffset(), + input.getLastOffset(), token.getErrorMessage()); } } return token; } +QPDFTokenizer::Token +QPDFTokenizer::readToken( + std::shared_ptr input, std::string const& context, bool allow_bad, size_t max_len) +{ + return readToken(*input, context, allow_bad, max_len); +} + bool QPDFTokenizer::nextToken(InputSource& input, std::string const& context, size_t max_len) { @@ -941,9 +927,8 @@ QPDFTokenizer::nextToken(InputSource& input, std::string const& context, size_t presentEOF(); if ((this->type == tt_eof) && (!this->allow_eof)) { - // Nothing in the qpdf library calls readToken - // without allowEOF anymore, so this case is not - // exercised. + // Nothing in the qpdf library calls readToken without allowEOF anymore, so this + // case is not exercised. this->type = tt_bad; this->error_message = "unexpected EOF"; offset = input.getLastOffset(); diff --git a/libqpdf/QPDF_Stream.cc b/libqpdf/QPDF_Stream.cc index 165b216f..45d6fb70 100644 --- a/libqpdf/QPDF_Stream.cc +++ b/libqpdf/QPDF_Stream.cc @@ -69,10 +69,9 @@ namespace } // namespace std::map QPDF_Stream::filter_abbreviations = { - // The PDF specification provides these filter abbreviations for - // use in inline images, but according to table H.1 in the pre-ISO - // versions of the PDF specification, Adobe Reader also accepts - // them for stream filters. + // The PDF specification provides these filter abbreviations for use in inline images, but + // according to table H.1 in the pre-ISO versions of the PDF specification, Adobe Reader also + // accepts them for stream filters. {"/AHx", "/ASCIIHexDecode"}, {"/A85", "/ASCII85Decode"}, {"/LZW", "/LZWDecode"}, @@ -118,8 +117,8 @@ QPDF_Stream::QPDF_Stream( length(length) { if (!stream_dict.isDictionary()) { - throw std::logic_error("stream object instantiated with non-dictionary " - "object for dictionary"); + throw std::logic_error( + "stream object instantiated with non-dictionary object for dictionary"); } auto descr = std::make_shared( qpdf->getFilename() + ", stream object " + og.unparse(' ')); @@ -198,18 +197,18 @@ QPDF_Stream::getStreamJSON( case qpdf_sj_none: case qpdf_sj_inline: if (p != nullptr) { - throw std::logic_error("QPDF_Stream::getStreamJSON: pipeline should " - "only be supplied when json_data is file"); + throw std::logic_error("QPDF_Stream::getStreamJSON: pipeline should only be supplied " + "when json_data is file"); } break; case qpdf_sj_file: if (p == nullptr) { - throw std::logic_error("QPDF_Stream::getStreamJSON: pipeline must " - "be supplied when json_data is file"); + throw std::logic_error( + "QPDF_Stream::getStreamJSON: pipeline must be supplied when json_data is file"); } if (data_filename.empty()) { - throw std::logic_error("QPDF_Stream::getStreamJSON: data_filename " - "must be supplied when json_data is file"); + throw std::logic_error("QPDF_Stream::getStreamJSON: data_filename must be supplied " + "when json_data is file"); } break; } @@ -244,8 +243,7 @@ QPDF_Stream::getStreamJSON( break; } } - // We can use unsafeShallowCopy because we are only - // touching top-level keys. + // We can use unsafeShallowCopy because we are only touching top-level keys. dict = this->stream_dict.unsafeShallowCopy(); dict.removeKey("/Length"); if (filter && filtered) { @@ -408,8 +406,7 @@ QPDF_Stream::filterable( return false; } - // filters now contains a list of filters to be applied in order. - // See which ones we can support. + // filters now contains a list of filters to be applied in order. See which ones we can support. // See if we can support any decode parameters that are specified. @@ -428,9 +425,8 @@ QPDF_Stream::filterable( } } - // Ignore /DecodeParms entirely if /Filters is empty. At least - // one case of a file whose /DecodeParms was [ << >> ] when - // /Filters was empty has been seen in the wild. + // Ignore /DecodeParms entirely if /Filters is empty. At least one case of a file whose + // /DecodeParms was [ << >> ] when /Filters was empty has been seen in the wild. if ((filters.size() != 0) && (decode_parms.size() != filters.size())) { warn("stream /DecodeParms length is inconsistent with filters"); filterable = false; @@ -502,9 +498,8 @@ QPDF_Stream::pipeStreamData( return filter; } - // Construct the pipeline in reverse order. Force pipelines we - // create to be deleted when this function finishes. Pipelines - // created by QPDFStreamFilter objects will be deleted by those + // Construct the pipeline in reverse order. Force pipelines we create to be deleted when this + // function finishes. Pipelines created by QPDFStreamFilter objects will be deleted by those // objects. std::vector> to_delete; @@ -568,8 +563,8 @@ QPDF_Stream::pipeStreamData( QTC::TC("qpdf", "QPDF_Stream pipe use stream provider"); } else { QTC::TC("qpdf", "QPDF_Stream provider length mismatch"); - // This would be caused by programmer error on the - // part of a library user, not by invalid input data. + // This would be caused by programmer error on the part of a library user, not by + // invalid input data. throw std::runtime_error( "stream data provider for " + og.unparse(' ') + " provided " + std::to_string(actual_length) + " bytes instead of expected " + @@ -602,14 +597,13 @@ QPDF_Stream::pipeStreamData( warn("content normalization encountered bad tokens"); if (normalizer->lastTokenWasBad()) { QTC::TC("qpdf", "QPDF_Stream bad token at end during normalize"); - warn("normalized content ended with a bad token; you may be able " - "to resolve this by coalescing content streams in combination " - "with normalizing content. From the command line, specify " - "--coalesce-contents"); + warn("normalized content ended with a bad token; you may be able to resolve this by " + "coalescing content streams in combination with normalizing content. From the " + "command line, specify --coalesce-contents"); } - warn("Resulting stream data may be corrupted but is may still useful " - "for manual inspection. For more information on this warning, " - "search for content normalization in the manual."); + warn("Resulting stream data may be corrupted but is may still useful for manual " + "inspection. For more information on this warning, search for content normalization " + "in the manual."); } return success; diff --git a/libqpdf/QPDF_encryption.cc b/libqpdf/QPDF_encryption.cc index 74136060..3fda99c4 100644 --- a/libqpdf/QPDF_encryption.cc +++ b/libqpdf/QPDF_encryption.cc @@ -137,9 +137,8 @@ pad_or_truncate_password_V4(std::string const& password, char k1[key_bytes]) void QPDF::trim_user_password(std::string& user_password) { - // Although unnecessary, this routine trims the padding string - // from the end of a user password. Its only purpose is for - // recovery of user passwords which is done in the test suite. + // Although unnecessary, this routine trims the padding string from the end of a user password. + // Its only purpose is for recovery of user passwords which is done in the test suite. char const* cstr = user_password.c_str(); size_t len = user_password.length(); if (len < key_bytes) { @@ -262,22 +261,17 @@ hash_V5( int round_number = 0; bool done = false; while (!done) { - // The hash algorithm has us setting K initially to the R5 - // value and then repeating a series of steps 64 times - // before starting with the termination case testing. The - // wording of the specification is very unclear as to the - // exact number of times it should be run since the - // wording about whether the initial setup counts as round - // 0 or not is ambiguous. This code counts the initial - // setup (R5) value as round 0, which appears to be - // correct. This was determined to be correct by - // increasing or decreasing the number of rounds by 1 or 2 - // from this value and generating 20 test files. In this - // interpretation, all the test files worked with Adobe - // Reader X. In the other configurations, many of the - // files did not work, and we were accurately able to - // predict which files didn't work by looking at the - // conditions under which we terminated repetition. + // The hash algorithm has us setting K initially to the R5 value and then repeating a + // series of steps 64 times before starting with the termination case testing. The + // wording of the specification is very unclear as to the exact number of times it + // should be run since the wording about whether the initial setup counts as round 0 or + // not is ambiguous. This code counts the initial setup (R5) value as round 0, which + // appears to be correct. This was determined to be correct by increasing or decreasing + // the number of rounds by 1 or 2 from this value and generating 20 test files. In this + // interpretation, all the test files worked with Adobe Reader X. In the other + // configurations, many of the files did not work, and we were accurately able to + // predict which files didn't work by looking at the conditions under which we + // terminated repetition. ++round_number; std::string K1 = password + K + udata; @@ -291,11 +285,10 @@ hash_V5( QUtil::unsigned_char_pointer(K.substr(16, 16)), 16); - // E_mod_3 is supposed to be mod 3 of the first 16 bytes - // of E taken as as a (128-bit) big-endian number. Since - // (xy mod n) is equal to ((x mod n) + (y mod n)) mod n - // and since 256 mod n is 1, we can just take the sums of - // the the mod 3s of each byte to get the same result. + // E_mod_3 is supposed to be mod 3 of the first 16 bytes of E taken as as a (128-bit) + // big-endian number. Since (xy mod n) is equal to ((x mod n) + (y mod n)) mod n and + // since 256 mod n is 1, we can just take the sums of the the mod 3s of each byte to get + // the same result. int E_mod_3 = 0; for (unsigned int i = 0; i < 16; ++i) { E_mod_3 += static_cast(E.at(i)); @@ -344,8 +337,7 @@ QPDF::compute_data_key( std::string result = encryption_key; if (encryption_V >= 5) { - // Algorithm 3.1a (PDF 1.7 extension level 3): just use - // encryption key straight. + // Algorithm 3.1a (PDF 1.7 extension level 3): just use encryption key straight. return result; } @@ -370,9 +362,8 @@ std::string QPDF::compute_encryption_key(std::string const& password, EncryptionData const& data) { if (data.getV() >= 5) { - // For V >= 5, the encryption key is generated and stored in - // the file, encrypted separately with both user and owner - // passwords. + // For V >= 5, the encryption key is generated and stored in the file, encrypted separately + // with both user and owner passwords. return recover_encryption_key_with_password(password, data); } else { // For V < 5, the encryption key is derived from the user @@ -386,12 +377,10 @@ QPDF::compute_encryption_key_from_password(std::string const& password, Encrypti { // Algorithm 3.2 from the PDF 1.7 Reference Manual - // This code does not properly handle Unicode passwords. - // Passwords are supposed to be converted from OS codepage - // characters to PDFDocEncoding. Unicode passwords are supposed - // to be converted to OS codepage before converting to - // PDFDocEncoding. We instead require the password to be - // presented in its final form. + // This code does not properly handle Unicode passwords. Passwords are supposed to be converted + // from OS codepage characters to PDFDocEncoding. Unicode passwords are supposed to be + // converted to OS codepage before converting to PDFDocEncoding. We instead require the + // password to be presented in its final form. MD5 md5; md5.encodeDataIncrementally(pad_or_truncate_password_V4(password).c_str(), key_bytes); @@ -681,11 +670,9 @@ QPDF::recover_encryption_key_with_password( { // Algorithm 3.2a from the PDF 1.7 extension level 3 - // This code does not handle Unicode passwords correctly. - // Empirical evidence suggests that most viewers don't. We are - // supposed to process the input string with the SASLprep (RFC - // 4013) profile of stringprep (RFC 3454) and then convert the - // result to UTF-8. + // This code does not handle Unicode passwords correctly. Empirical evidence suggests that most + // viewers don't. We are supposed to process the input string with the SASLprep (RFC 4013) + // profile of stringprep (RFC 3454) and then convert the result to UTF-8. perms_valid = false; std::string key_password = truncate_password_V5(password); @@ -738,18 +725,16 @@ QPDF::initializeEncryption() } m->encp->encryption_initialized = true; - // After we initialize encryption parameters, we must used stored - // key information and never look at /Encrypt again. Otherwise, - // things could go wrong if someone mutates the encryption + // After we initialize encryption parameters, we must used stored key information and never look + // at /Encrypt again. Otherwise, things could go wrong if someone mutates the encryption // dictionary. if (!m->trailer.hasKey("/Encrypt")) { return; } - // Go ahead and set m->encrypted here. That way, isEncrypted - // will return true even if there were errors reading the - // encryption dictionary. + // Go ahead and set m->encrypted here. That way, isEncrypted will return true even if there + // were errors reading the encryption dictionary. m->encp->encrypted = true; std::string id1; @@ -757,9 +742,8 @@ QPDF::initializeEncryption() if ((id_obj.isArray() && (id_obj.getArrayNItems() == 2) && id_obj.getArrayItem(0).isString())) { id1 = id_obj.getArrayItem(0).getStringValue(); } else { - // Treating a missing ID as the empty string enables qpdf to - // decrypt some invalid encrypted files with no /ID that - // poppler can read but Adobe Reader can't. + // Treating a missing ID as the empty string enables qpdf to decrypt some invalid encrypted + // files with no /ID that poppler can read but Adobe Reader can't. warn(damagedPDF("trailer", "invalid /ID in trailer dictionary")); } @@ -800,8 +784,8 @@ QPDF::initializeEncryption() std::string U = encryption_dict.getKey("/U").getStringValue(); int P = static_cast(encryption_dict.getKey("/P").getIntValue()); - // If supporting new encryption R/V values, remember to update - // error message inside this if statement. + // If supporting new encryption R/V values, remember to update error message inside this if + // statement. if (!(((R >= 2) && (R <= 6)) && ((V == 1) || (V == 2) || (V == 4) || (V == 5)))) { throw QPDFExc( qpdf_e_unsupported, @@ -893,8 +877,7 @@ QPDF::initializeEncryption() QTC::TC("qpdf", "QPDF_encryption CFM AESV3"); method = e_aesv3; } else { - // Don't complain now -- maybe we won't need - // to reference this type. + // Don't complain now -- maybe we won't need to reference this type. method = e_unknown; } } @@ -908,20 +891,15 @@ QPDF::initializeEncryption() m->encp->cf_stream = interpretCF(m->encp, StmF); m->encp->cf_string = interpretCF(m->encp, StrF); if (EFF.isName()) { - // qpdf does not use this for anything other than - // informational purposes. This is intended to instruct - // conforming writers on which crypt filter should be used - // when new file attachments are added to a PDF file, but - // qpdf never generates encrypted files with non-default - // crypt filters. Prior to 10.2, I was under the mistaken - // impression that this was supposed to be used for - // decrypting attachments, but the code was wrong in a way - // that turns out not to have mattered because no writers - // were generating files the way I was imagining. Still, - // providing this information could be useful when looking - // at a file generated by something else, such as Acrobat - // when specifying that only attachments should be - // encrypted. + // qpdf does not use this for anything other than informational purposes. This is + // intended to instruct conforming writers on which crypt filter should be used when new + // file attachments are added to a PDF file, but qpdf never generates encrypted files + // with non-default crypt filters. Prior to 10.2, I was under the mistaken impression + // that this was supposed to be used for decrypting attachments, but the code was wrong + // in a way that turns out not to have mattered because no writers were generating files + // the way I was imagining. Still, providing this information could be useful when + // looking at a file generated by something else, such as Acrobat when specifying that + // only attachments should be encrypted. m->encp->cf_file = interpretCF(m->encp, EFF); } else { m->encp->cf_file = m->encp->cf_stream; @@ -935,8 +913,7 @@ QPDF::initializeEncryption() m->encp->owner_password_matched = check_owner_password(m->encp->user_password, m->encp->provided_password, data); if (m->encp->owner_password_matched && (V < 5)) { - // password supplied was owner password; user_password has - // been initialized for V < 5 + // password supplied was owner password; user_password has been initialized for V < 5 if (getTrimmedUserPassword() == m->encp->provided_password) { m->encp->user_password_matched = true; QTC::TC("qpdf", "QPDF_encryption user matches owner V < 5"); @@ -958,14 +935,12 @@ QPDF::initializeEncryption() if (m->provided_password_is_hex_key) { m->encp->encryption_key = QUtil::hex_decode(m->encp->provided_password); } else if (V < 5) { - // For V < 5, the user password is encrypted with the owner - // password, and the user password is always used for - // computing the encryption key. + // For V < 5, the user password is encrypted with the owner password, and the user password + // is always used for computing the encryption key. m->encp->encryption_key = compute_encryption_key(m->encp->user_password, data); } else { - // For V >= 5, either password can be used independently to - // compute the encryption key, and neither password can be - // used to recover the other. + // For V >= 5, either password can be used independently to compute the encryption key, and + // neither password can be used to recover the other. bool perms_valid; m->encp->encryption_key = recover_encryption_key_with_password(m->encp->provided_password, data, perms_valid); @@ -1026,8 +1001,7 @@ QPDF::decryptString(std::string& str, QPDFObjGen const& og) default: warn(damagedPDF("unknown encryption filter for strings (check /StrF in " "/Encrypt dictionary); strings may be decrypted improperly")); - // To avoid repeated warnings, reset cf_string. Assume - // we'd want to use AES if V == 4. + // To avoid repeated warnings, reset cf_string. Assume we'd want to use AES if V == 4. m->encp->cf_string = e_aes; use_aes = true; break; @@ -1052,8 +1026,8 @@ QPDF::decryptString(std::string& str, QPDFObjGen const& og) } else { QTC::TC("qpdf", "QPDF_encryption rc4 decode string"); size_t vlen = str.length(); - // Using std::shared_ptr guarantees that tmp will - // be freed even if rc4.process throws an exception. + // Using std::shared_ptr guarantees that tmp will be freed even if rc4.process throws an + // exception. auto tmp = QUtil::make_unique_cstr(str); RC4 rc4(QUtil::unsigned_char_pointer(key), toI(key.length())); auto data = QUtil::unsigned_char_pointer(tmp.get()); @@ -1154,8 +1128,7 @@ QPDF::decryptStream( file->getLastOffset(), "unknown encryption filter for streams (check " + method_source + "); streams may be decrypted improperly")); - // To avoid repeated warnings, reset cf_stream. Assume - // we'd want to use AES if V == 4. + // To avoid repeated warnings, reset cf_stream. Assume we'd want to use AES if V == 4. encp->cf_stream = e_aes; use_aes = true; break; diff --git a/libqpdf/QPDF_json.cc b/libqpdf/QPDF_json.cc index 66d4b314..c74cf4f7 100644 --- a/libqpdf/QPDF_json.cc +++ b/libqpdf/QPDF_json.cc @@ -12,8 +12,7 @@ #include #include -// This chart shows an example of the state transitions that would -// occur in parsing a minimal file. +// This chart shows an example of the state transitions that would occur in parsing a minimal file. // | st_initial // { | -> st_top @@ -414,9 +413,9 @@ QPDF::JSONReactor::containerEnd(JSON const& value) object_stack.pop_back(); } } else if ((state == st_top) && (from_state == st_qpdf)) { - // Handle dangling indirect object references which the PDF spec says to - // treat as nulls. It's tempting to make this an error, but that would - // be wrong since valid input files may have these. + // Handle dangling indirect object references which the PDF spec says to treat as nulls. + // It's tempting to make this an error, but that would be wrong since valid input files may + // have these. for (auto& oc: pdf.m->obj_cache) { if (oc.second.object->getTypeCode() == ::ot_reserved && reserved.count(oc.first) == 0) { QTC::TC("qpdf", "QPDF_json non-trivial null reserved"); @@ -446,8 +445,7 @@ QPDF::JSONReactor::topLevelScalar() void QPDF::JSONReactor::nestedState(std::string const& key, JSON const& value, state_e next) { - // Use this method when the next state is for processing a nested - // dictionary. + // Use this method when the next state is for processing a nested dictionary. if (value.isDictionary()) { this->next_state = next; } else { @@ -531,8 +529,8 @@ QPDF::JSONReactor::dictionaryItem(std::string const& key, JSON const& value) error(value.getStart(), "calledgetallpages must be a boolean"); } } else { - // ignore unknown keys for forward compatibility and to - // skip keys we don't care about like "maxobjectid". + // ignore unknown keys for forward compatibility and to skip keys we don't care about + // like "maxobjectid". QTC::TC("qpdf", "QPDF_json ignore second-level key"); next_state = st_ignore; } @@ -594,8 +592,7 @@ QPDF::JSONReactor::dictionaryItem(std::string const& key, JSON const& value) this->pdf.m->trailer = makeObject(value); setObjectDescription(this->pdf.m->trailer, value); } else if (key == "stream") { - // Don't need to set saw_stream here since there's already - // an error. + // Don't need to set saw_stream here since there's already an error. QTC::TC("qpdf", "QPDF_json trailer stream"); error(value.getStart(), "the trailer may not be a stream"); next_state = st_ignore; @@ -616,8 +613,8 @@ QPDF::JSONReactor::dictionaryItem(std::string const& key, JSON const& value) auto uninitialized = QPDFObjectHandle(); if (key == "dict") { this->saw_dict = true; - // Since a stream dictionary must be a dictionary, we can - // use nestedState to transition to st_value. + // Since a stream dictionary must be a dictionary, we can use nestedState to transition + // to st_value. nestedState("stream.dict", value, st_object); auto dict = makeObject(value); if (dict.isDictionary()) { diff --git a/libqpdf/QPDF_linearization.cc b/libqpdf/QPDF_linearization.cc index 65357b9a..faebf5b6 100644 --- a/libqpdf/QPDF_linearization.cc +++ b/libqpdf/QPDF_linearization.cc @@ -22,8 +22,8 @@ load_vector_int( BitStream& bit_stream, int nitems, std::vector& vec, int bits_wanted, int_type T::*field) { bool append = vec.empty(); - // nitems times, read bits_wanted from the given bit stream, - // storing results in the ith vector entry. + // nitems times, read bits_wanted from the given bit stream, storing results in the ith vector + // entry. for (size_t i = 0; i < QIntC::to_size(nitems); ++i) { if (append) { @@ -34,8 +34,8 @@ load_vector_int( if (QIntC::to_int(vec.size()) != nitems) { throw std::logic_error("vector has wrong size in load_vector_int"); } - // The PDF spec says that each hint table starts at a byte - // boundary. Each "row" actually must start on a byte boundary. + // The PDF spec says that each hint table starts at a byte boundary. Each "row" actually must + // start on a byte boundary. bit_stream.skipToNextByte(); } @@ -49,8 +49,8 @@ load_vector_vector( int bits_wanted, std::vector T::*vec2) { - // nitems1 times, read nitems2 (from the ith element of vec1) items - // into the vec2 vector field of the ith item of vec1. + // nitems1 times, read nitems2 (from the ith element of vec1) items into the vec2 vector field + // of the ith item of vec1. for (size_t i1 = 0; i1 < QIntC::to_size(nitems1); ++i1) { for (int i2 = 0; i2 < vec1.at(i1).*nitems2; ++i2) { (vec1.at(i1).*vec2).push_back(bit_stream.getBitsInt(QIntC::to_size(bits_wanted))); @@ -83,18 +83,15 @@ QPDF::checkLinearization() bool QPDF::isLinearized() { - // If the first object in the file is a dictionary with a suitable - // /Linearized key and has an /L key that accurately indicates the - // file size, initialize m->lindict and return true. - - // A linearized PDF spec's first object will be contained within - // the first 1024 bytes of the file and will be a dictionary with - // a valid /Linearized key. This routine looks for that and does - // no additional validation. - - // The PDF spec says the linearization dictionary must be - // completely contained within the first 1024 bytes of the file. - // Add a byte for a null terminator. + // If the first object in the file is a dictionary with a suitable /Linearized key and has an /L + // key that accurately indicates the file size, initialize m->lindict and return true. + + // A linearized PDF spec's first object will be contained within the first 1024 bytes of the + // file and will be a dictionary with a valid /Linearized key. This routine looks for that and + // does no additional validation. + + // The PDF spec says the linearization dictionary must be completely contained within the first + // 1024 bytes of the file. Add a byte for a null terminator. static int const tbuf_size = 1025; auto b = std::make_unique(tbuf_size); @@ -161,8 +158,8 @@ QPDF::isLinearized() void QPDF::readLinearizationData() { - // This function throws an exception (which is trapped by - // checkLinearization()) for any errors that prevent loading. + // This function throws an exception (which is trapped by checkLinearization()) for any errors + // that prevent loading. if (!isLinearized()) { throw std::logic_error("called readLinearizationData for file" @@ -206,8 +203,8 @@ QPDF::readLinearizationData() int H1_offset = 0; int H1_length = 0; if (H_items.size() == 4) { - // Acrobat doesn't read or write these (as PDF 1.4), so we - // don't have a way to generate a test case. + // Acrobat doesn't read or write these (as PDF 1.4), so we don't have a way to generate a + // test case. // QTC::TC("qpdf", "QPDF overflow hint table"); H1_offset = H_items.at(2); H1_length = H_items.at(3); @@ -224,9 +221,8 @@ QPDF::readLinearizationData() // Store linearization parameter data - // Various places in the code use linp.npages, which is - // initialized from N, to pre-allocate memory, so make sure it's - // accurate and bail right now if it's not. + // Various places in the code use linp.npages, which is initialized from N, to pre-allocate + // memory, so make sure it's accurate and bail right now if it's not. if (N.getIntValue() != static_cast(getAllPages().size())) { throw damagedPDF("linearization hint table", "/N does not match number of pages"); } @@ -299,11 +295,10 @@ QPDF::readHintStream(Pipeline& pl, qpdf_offset_t offset, size_t length) QPDFObjectHandle Hdict = H.getDict(); - // Some versions of Acrobat make /Length indirect and place it - // immediately after the stream, increasing length to cover it, - // even though the specification says all objects in the - // linearization parameter dictionary must be direct. We have to - // get the file position of the end of length in this case. + // Some versions of Acrobat make /Length indirect and place it immediately after the stream, + // increasing length to cover it, even though the specification says all objects in the + // linearization parameter dictionary must be direct. We have to get the file position of the + // end of length in this case. QPDFObjectHandle length_obj = Hdict.getKey("/Length"); if (length_obj.isIndirect()) { QTC::TC("qpdf", "QPDF hint table length indirect"); @@ -329,8 +324,7 @@ QPDF::readHintStream(Pipeline& pl, qpdf_offset_t offset, size_t length) void QPDF::readHPageOffset(BitStream h) { - // All comments referring to the PDF spec refer to the spec for - // version 1.4. + // All comments referring to the PDF spec refer to the spec for version 1.4. HPageOffset& t = m->page_offset_hints; @@ -402,9 +396,8 @@ QPDF::readHSharedObject(BitStream h) load_vector_int(h, nitems, entries, 1, &HSharedObjectEntry::signature_present); for (size_t i = 0; i < toS(nitems); ++i) { if (entries.at(i).signature_present) { - // Skip 128-bit MD5 hash. These are not supported by - // acrobat, so they should probably never be there. We - // have no test case for this. + // Skip 128-bit MD5 hash. These are not supported by acrobat, so they should probably + // never be there. We have no test case for this. for (int j = 0; j < 4; ++j) { (void)h.getBits(32); } @@ -425,8 +418,7 @@ QPDF::readHGeneric(BitStream h, HGeneric& t) bool QPDF::checkLinearizationInternal() { - // All comments referring to the PDF spec refer to the spec for - // version 1.4. + // All comments referring to the PDF spec refer to the spec for version 1.4. // Check all values in linearization parameter dictionary @@ -476,24 +468,21 @@ QPDF::checkLinearizationInternal() "; file = " + std::to_string(m->file->tell())); } - // P: first page number -- Implementation note 124 says Acrobat - // ignores this value, so we will too. + // P: first page number -- Implementation note 124 says Acrobat ignores this value, so we will + // too. - // Check numbering of compressed objects in each xref section. - // For linearized files, all compressed objects are supposed to be - // at the end of the containing xref section if any object streams - // are in use. + // Check numbering of compressed objects in each xref section. For linearized files, all + // compressed objects are supposed to be at the end of the containing xref section if any object + // streams are in use. if (m->uncompressed_after_compressed) { - linearizationWarning("linearized file contains an uncompressed object" - " after a compressed one in a cross-reference stream"); + linearizationWarning("linearized file contains an uncompressed object after a compressed " + "one in a cross-reference stream"); } - // Further checking requires optimization and order calculation. - // Don't allow optimization to make changes. If it has to, then - // the file is not properly linearized. We use the xref table to - // figure out which objects are compressed and which are - // uncompressed. + // Further checking requires optimization and order calculation. Don't allow optimization to + // make changes. If it has to, then the file is not properly linearized. We use the xref table + // to figure out which objects are compressed and which are uncompressed. { // local scope std::map object_stream_data; for (auto const& iter: m->xref_table) { @@ -507,16 +496,13 @@ QPDF::checkLinearizationInternal() calculateLinearizationData(object_stream_data); } - // E: offset of end of first page -- Implementation note 123 says - // Acrobat includes on extra object here by mistake. pdlin fails - // to place thumbnail images in section 9, so when thumbnails are - // present, it also gets the wrong value for /E. It also doesn't - // count outlines here when it should even though it places them - // in part 6. This code fails to put thread information - // dictionaries in part 9, so it actually gets the wrong value for - // E when threads are present. In that case, it would probably - // agree with pdlin. As of this writing, the test suite doesn't - // contain any files with threads. + // E: offset of end of first page -- Implementation note 123 says Acrobat includes on extra + // object here by mistake. pdlin fails to place thumbnail images in section 9, so when + // thumbnails are present, it also gets the wrong value for /E. It also doesn't count outlines + // here when it should even though it places them in part 6. This code fails to put thread + // information dictionaries in part 9, so it actually gets the wrong value for E when threads + // are present. In that case, it would probably agree with pdlin. As of this writing, the test + // suite doesn't contain any files with threads. if (m->part6.empty()) { stopOnError("linearization part 6 unexpectedly empty"); @@ -577,8 +563,7 @@ QPDF::getLinearizationOffset(QPDFObjGen const& og) break; case 2: - // For compressed objects, return the offset of the object - // stream that contains them. + // For compressed objects, return the offset of the object stream that contains them. result = getLinearizationOffset(QPDFObjGen(entry.getObjStreamNumber(), 0)); break; @@ -611,8 +596,7 @@ QPDF::lengthNextN(int first_object, int n) "no xref table entry for " + std::to_string(first_object + i) + " 0"); } else { if (m->obj_cache.count(og) == 0) { - stopOnError("found unknown object while" - " calculating length for linearization data"); + stopOnError("found unknown object while calculating length for linearization data"); } length += toI(m->obj_cache[og].end_after_space - getLinearizationOffset(og)); } @@ -624,22 +608,17 @@ void QPDF::checkHPageOffset( std::vector const& pages, std::map& shared_idx_to_obj) { - // Implementation note 126 says Acrobat always sets - // delta_content_offset and delta_content_length in the page - // offset header dictionary to 0. It also states that - // min_content_offset in the per-page information is always 0, - // which is an incorrect value. - - // Implementation note 127 explains that Acrobat always sets item - // 8 (min_content_length) to zero, item 9 - // (nbits_delta_content_length) to the value of item 5 - // (nbits_delta_page_length), and item 7 of each per-page hint - // table (delta_content_length) to item 2 (delta_page_length) of - // that entry. Acrobat ignores these values when reading files. - - // Empirically, it also seems that Acrobat sometimes puts items - // under a page's /Resources dictionary in with shared objects - // even when they are private. + // Implementation note 126 says Acrobat always sets delta_content_offset and + // delta_content_length in the page offset header dictionary to 0. It also states that + // min_content_offset in the per-page information is always 0, which is an incorrect value. + + // Implementation note 127 explains that Acrobat always sets item 8 (min_content_length) to + // zero, item 9 (nbits_delta_content_length) to the value of item 5 (nbits_delta_page_length), + // and item 7 of each per-page hint table (delta_content_length) to item 2 (delta_page_length) + // of that entry. Acrobat ignores these values when reading files. + + // Empirically, it also seems that Acrobat sometimes puts items under a page's /Resources + // dictionary in with shared objects even when they are private. int npages = toI(pages.size()); qpdf_offset_t table_offset = adjusted_offset(m->page_offset_hints.first_page_offset); @@ -670,13 +649,12 @@ QPDF::checkHPageOffset( std::to_string(h_nobjects) + "; computed = " + std::to_string(ce.nobjects)); } - // Use value for number of objects in hint table rather than - // computed value if there is a discrepancy. + // Use value for number of objects in hint table rather than computed value if there is a + // discrepancy. int length = lengthNextN(first_object, h_nobjects); int h_length = toI(he.delta_page_length + m->page_offset_hints.min_page_length); if (length != h_length) { - // This condition almost certainly indicates a bad hint - // table or a bug in this code. + // This condition almost certainly indicates a bad hint table or a bug in this code. linearizationWarning( "page length mismatch for page " + std::to_string(pageno) + ": hint table = " + std::to_string(h_length) + "; computed length = " + std::to_string(length) + @@ -690,8 +668,8 @@ QPDF::checkHPageOffset( std::set computed_shared; if ((pageno == 0) && (he.nshared_objects > 0)) { - // pdlin and Acrobat both do this even though the spec - // states clearly and unambiguously that they should not. + // pdlin and Acrobat both do this even though the spec states clearly and unambiguously + // that they should not. linearizationWarning("page 0 has shared identifier entries"); } @@ -724,9 +702,8 @@ QPDF::checkHPageOffset( for (int iter: computed_shared) { if (!hint_shared.count(iter)) { - // Acrobat does not put some things including at least - // built-in fonts and procsets here, at least in some - // cases. + // Acrobat does not put some things including at least built-in fonts and procsets + // here, at least in some cases. linearizationWarning( ("page " + std::to_string(pageno) + ": shared object " + std::to_string(iter) + ": in computed list but not hint table")); @@ -738,31 +715,26 @@ QPDF::checkHPageOffset( void QPDF::checkHSharedObject(std::vector const& pages, std::map& idx_to_obj) { - // Implementation note 125 says shared object groups always - // contain only one object. Implementation note 128 says that - // Acrobat always nbits_nobjects to zero. Implementation note 130 - // says that Acrobat does not support more than one shared object - // per group. These are all consistent. + // Implementation note 125 says shared object groups always contain only one object. + // Implementation note 128 says that Acrobat always nbits_nobjects to zero. Implementation note + // 130 says that Acrobat does not support more than one shared object per group. These are all + // consistent. - // Implementation note 129 states that MD5 signatures are not - // implemented in Acrobat, so signature_present must always be - // zero. + // Implementation note 129 states that MD5 signatures are not implemented in Acrobat, so + // signature_present must always be zero. - // Implementation note 131 states that first_shared_obj and - // first_shared_offset have meaningless values for single-page - // files. + // Implementation note 131 states that first_shared_obj and first_shared_offset have meaningless + // values for single-page files. - // Empirically, Acrobat and pdlin generate incorrect values for - // these whenever there are no shared objects not referenced by - // the first page (i.e., nshared_total == nshared_first_page). + // Empirically, Acrobat and pdlin generate incorrect values for these whenever there are no + // shared objects not referenced by the first page (i.e., nshared_total == nshared_first_page). HSharedObject& so = m->shared_object_hints; if (so.nshared_total < so.nshared_first_page) { linearizationWarning("shared object hint table: ntotal < nfirst_page"); } else { - // The first nshared_first_page objects are consecutive - // objects starting with the first page object. The rest are - // consecutive starting from the first_shared_obj object. + // The first nshared_first_page objects are consecutive objects starting with the first page + // object. The rest are consecutive starting from the first_shared_obj object. int cur_object = pages.at(0).getObjectID(); for (int i = 0; i < so.nshared_total; ++i) { if (i == so.nshared_first_page) { @@ -814,12 +786,10 @@ QPDF::checkHSharedObject(std::vector const& pages, std::mapc_outline_data.nobjects == m->outline_hints.nobjects) { @@ -831,9 +801,8 @@ QPDF::checkHOutlines() // Check length and offset. Acrobat gets these wrong. QPDFObjectHandle outlines = getRoot().getKey("/Outlines"); if (!outlines.isIndirect()) { - // This case is not exercised in test suite since not - // permitted by the spec, but if this does occur, the - // code below would fail. + // This case is not exercised in test suite since not permitted by the spec, but if + // this does occur, the code below would fail. linearizationWarning("/Outlines key of root dictionary is not indirect"); return; } @@ -906,9 +875,8 @@ QPDF::dumpLinearizationDataInternal() qpdf_offset_t QPDF::adjusted_offset(qpdf_offset_t offset) { - // All offsets >= H_offset have to be increased by H_length - // since all hint table location values disregard the hint table - // itself. + // All offsets >= H_offset have to be increased by H_length since all hint table location values + // disregard the hint table itself. if (offset >= m->linp.H_offset) { return offset + m->linp.H_length; } @@ -971,8 +939,8 @@ QPDF::dumpHSharedObject() *m->log->getInfo() << "Shared Object " << i << ":\n" << " group length: " << se.delta_group_length + t.min_group_length << "\n"; - // PDF spec says signature present nobjects_minus_one are - // always 0, so print them only if they have a non-zero value. + // PDF spec says signature present nobjects_minus_one are always 0, so print them only if + // they have a non-zero value. if (se.signature_present) { *m->log->getInfo() << " signature present\n"; } @@ -994,44 +962,38 @@ QPDF::dumpHGeneric(HGeneric& t) void QPDF::calculateLinearizationData(std::map const& object_stream_data) { - // This function calculates the ordering of objects, divides them - // into the appropriate parts, and computes some values for the - // linearization parameter dictionary and hint tables. The file - // must be optimized (via calling optimize()) prior to calling - // this function. Note that actual offsets and lengths are not - // computed here, but anything related to object ordering is. + // This function calculates the ordering of objects, divides them into the appropriate parts, + // and computes some values for the linearization parameter dictionary and hint tables. The + // file must be optimized (via calling optimize()) prior to calling this function. Note that + // actual offsets and lengths are not computed here, but anything related to object ordering is. if (m->object_to_obj_users.empty()) { - // Note that we can't call optimize here because we don't know - // whether it should be called with or without allow changes. - throw std::logic_error("INTERNAL ERROR: QPDF::calculateLinearizationData " - "called before optimize()"); + // Note that we can't call optimize here because we don't know whether it should be called + // with or without allow changes. + throw std::logic_error( + "INTERNAL ERROR: QPDF::calculateLinearizationData called before optimize()"); } - // Separate objects into the categories sufficient for us to - // determine which part of the linearized file should contain the - // object. This categorization is useful for other purposes as - // well. Part numbers refer to version 1.4 of the PDF spec. + // Separate objects into the categories sufficient for us to determine which part of the + // linearized file should contain the object. This categorization is useful for other purposes + // as well. Part numbers refer to version 1.4 of the PDF spec. - // Parts 1, 3, 5, 10, and 11 don't contain any objects from the - // original file (except the trailer dictionary in part 11). + // Parts 1, 3, 5, 10, and 11 don't contain any objects from the original file (except the + // trailer dictionary in part 11). - // Part 4 is the document catalog (root) and the following root - // keys: /ViewerPreferences, /PageMode, /Threads, /OpenAction, - // /AcroForm, /Encrypt. Note that Thread information dictionaries - // are supposed to appear in part 9, but we are disregarding that - // recommendation for now. + // Part 4 is the document catalog (root) and the following root keys: /ViewerPreferences, + // /PageMode, /Threads, /OpenAction, /AcroForm, /Encrypt. Note that Thread information + // dictionaries are supposed to appear in part 9, but we are disregarding that recommendation + // for now. - // Part 6 is the first page section. It includes all remaining - // objects referenced by the first page including shared objects - // but not including thumbnails. Additionally, if /PageMode is + // Part 6 is the first page section. It includes all remaining objects referenced by the first + // page including shared objects but not including thumbnails. Additionally, if /PageMode is // /Outlines, then information from /Outlines also appears here. - // Part 7 contains remaining objects private to pages other than - // the first page. + // Part 7 contains remaining objects private to pages other than the first page. - // Part 8 contains all remaining shared objects except those that - // are shared only within thumbnails. + // Part 8 contains all remaining shared objects except those that are shared only within + // thumbnails. // Part 9 contains all remaining objects. @@ -1176,42 +1138,35 @@ QPDF::calculateLinearizationData(std::map const& object_stream_data) } } - // Generate ordering for objects in the output file. Sometimes we - // just dump right from a set into a vector. Rather than - // optimizing this by going straight into the vector, we'll leave - // these phases separate for now. That way, this section can be - // concerned only with ordering, and the above section can be - // considered only with categorization. Note that sets of - // QPDFObjGens are sorted by QPDFObjGen. In a linearized file, - // objects appear in sequence with the possible exception of hints - // tables which we won't see here anyway. That means that running - // calculateLinearizationData() on a linearized file should give - // results identical to the original file ordering. - - // We seem to traverse the page tree a lot in this code, but we - // can address this for a future code optimization if necessary. - // Premature optimization is the root of all evil. + // Generate ordering for objects in the output file. Sometimes we just dump right from a set + // into a vector. Rather than optimizing this by going straight into the vector, we'll leave + // these phases separate for now. That way, this section can be concerned only with ordering, + // and the above section can be considered only with categorization. Note that sets of + // QPDFObjGens are sorted by QPDFObjGen. In a linearized file, objects appear in sequence with + // the possible exception of hints tables which we won't see here anyway. That means that + // running calculateLinearizationData() on a linearized file should give results identical to + // the original file ordering. + + // We seem to traverse the page tree a lot in this code, but we can address this for a future + // code optimization if necessary. Premature optimization is the root of all evil. std::vector pages; { // local scope - // Map all page objects to the containing object stream. This - // should be a no-op in a properly linearized file. + // Map all page objects to the containing object stream. This should be a no-op in a + // properly linearized file. for (auto oh: getAllPages()) { pages.push_back(getUncompressedObject(oh, object_stream_data)); } } int npages = toI(pages.size()); - // We will be initializing some values of the computed hint - // tables. Specifically, we can initialize any items that deal - // with object numbers or counts but not any items that deal with - // lengths or offsets. The code that writes linearized files will - // have to fill in these values during the first pass. The - // validation code can compute them relatively easily given the - // rest of the information. - - // npages is the size of the existing pages vector, which has been - // created by traversing the pages tree, and as such is a - // reasonable size. + // We will be initializing some values of the computed hint tables. Specifically, we can + // initialize any items that deal with object numbers or counts but not any items that deal with + // lengths or offsets. The code that writes linearized files will have to fill in these values + // during the first pass. The validation code can compute them relatively easily given the rest + // of the information. + + // npages is the size of the existing pages vector, which has been created by traversing the + // pages tree, and as such is a reasonable size. m->c_linp.npages = npages; m->c_page_offset_data.entries = std::vector(toS(npages)); @@ -1226,11 +1181,9 @@ QPDF::calculateLinearizationData(std::map const& object_stream_data) m->part4.push_back(getObject(og)); } - // Part 6: first page objects. Note: implementation note 124 - // states that Acrobat always treats page 0 as the first page for - // linearization regardless of /OpenAction. pdlin doesn't provide - // any option to set this and also disregards /OpenAction. We - // will do the same. + // Part 6: first page objects. Note: implementation note 124 states that Acrobat always treats + // page 0 as the first page for linearization regardless of /OpenAction. pdlin doesn't provide + // any option to set this and also disregards /OpenAction. We will do the same. // First, place the actual first page object itself. if (pages.empty()) { @@ -1245,10 +1198,9 @@ QPDF::calculateLinearizationData(std::map const& object_stream_data) m->c_linp.first_page_object = pages.at(0).getObjectID(); m->part6.push_back(pages.at(0)); - // The PDF spec "recommends" an order for the rest of the objects, - // but we are going to disregard it except to the extent that it - // groups private and shared objects contiguously for the sake of - // hint tables. + // The PDF spec "recommends" an order for the rest of the objects, but we are going to disregard + // it except to the extent that it groups private and shared objects contiguously for the sake + // of hint tables. for (auto const& og: lc_first_page_private) { m->part6.push_back(getObject(og)); @@ -1263,11 +1215,9 @@ QPDF::calculateLinearizationData(std::map const& object_stream_data) pushOutlinesToPart(m->part6, lc_outlines, object_stream_data); } - // Fill in page offset hint table information for the first page. - // The PDF spec says that nshared_objects should be zero for the - // first page. pdlin does not appear to obey this, but it fills - // in garbage values for all the shared object identifiers on the - // first page. + // Fill in page offset hint table information for the first page. The PDF spec says that + // nshared_objects should be zero for the first page. pdlin does not appear to obey this, but + // it fills in garbage values for all the shared object identifiers on the first page. m->c_page_offset_data.entries.at(0).nobjects = toI(m->part6.size()); @@ -1287,8 +1237,8 @@ QPDF::calculateLinearizationData(std::map const& object_stream_data) lc_other_page_private.erase(page_og); m->part7.push_back(pages.at(i)); - // Place all non-shared objects referenced by this page, - // updating the page object count for the hint table. + // Place all non-shared objects referenced by this page, updating the page object count for + // the hint table. m->c_page_offset_data.entries.at(i).nobjects = 1; @@ -1321,12 +1271,10 @@ QPDF::calculateLinearizationData(std::map const& object_stream_data) // Part 9: other objects - // The PDF specification makes recommendations on ordering here. - // We follow them only to a limited extent. Specifically, we put - // the pages tree first, then private thumbnail objects in page - // order, then shared thumbnail objects, and then outlines (unless - // in part 6). After that, we throw all remaining objects in - // arbitrary order. + // The PDF specification makes recommendations on ordering here. We follow them only to a + // limited extent. Specifically, we put the pages tree first, then private thumbnail objects in + // page order, then shared thumbnail objects, and then outlines (unless in part 6). After that, + // we throw all remaining objects in arbitrary order. // Place the pages tree. std::set pages_ogs = @@ -1342,9 +1290,8 @@ QPDF::calculateLinearizationData(std::map const& object_stream_data) } } - // Place private thumbnail images in page order. Slightly more - // information would be required if we were going to bother with - // thumbnail hint tables. + // Place private thumbnail images in page order. Slightly more information would be required if + // we were going to bother with thumbnail hint tables. for (size_t i = 0; i < toS(npages); ++i) { QPDFObjectHandle thumb = pages.at(i).getKey("/Thumb"); thumb = getUncompressedObject(thumb, object_stream_data); @@ -1355,11 +1302,9 @@ QPDF::calculateLinearizationData(std::map const& object_stream_data) lc_thumbnail_private.erase(thumb_og); m->part9.push_back(thumb); } else { - // No internal error this time...there's nothing to - // stop this object from having been referred to - // somewhere else outside of a page's /Thumb, and if - // it had been, there's nothing to prevent it from - // having been in some set other than + // No internal error this time...there's nothing to stop this object from having + // been referred to somewhere else outside of a page's /Thumb, and if it had been, + // there's nothing to prevent it from having been in some set other than // lc_thumbnail_private. } std::set& ogs = m->obj_user_to_objects[ObjUser(ObjUser::ou_thumb, toI(i))]; @@ -1372,9 +1317,8 @@ QPDF::calculateLinearizationData(std::map const& object_stream_data) } } if (!lc_thumbnail_private.empty()) { - stopOnError("INTERNAL ERROR: " - "QPDF::calculateLinearizationData: lc_thumbnail_private " - "not empty after placing thumbnails"); + stopOnError("INTERNAL ERROR: QPDF::calculateLinearizationData: lc_thumbnail_private not " + "empty after placing thumbnails"); } // Place shared thumbnail objects @@ -1404,17 +1348,15 @@ QPDF::calculateLinearizationData(std::map const& object_stream_data) std::to_string(num_placed) + "; number of objects: " + std::to_string(num_wanted)); } - // Calculate shared object hint table information including - // references to shared objects from page offset hint data. + // Calculate shared object hint table information including references to shared objects from + // page offset hint data. - // The shared object hint table consists of all part 6 (whether - // shared or not) in order followed by all part 8 objects in - // order. Add the objects to shared object data keeping a map of - // object number to index. Then populate the shared object - // information for the pages. + // The shared object hint table consists of all part 6 (whether shared or not) in order followed + // by all part 8 objects in order. Add the objects to shared object data keeping a map of + // object number to index. Then populate the shared object information for the pages. - // Note that two objects never have the same object number, so we - // can map from object number only without regards to generation. + // Note that two objects never have the same object number, so we can map from object number + // only without regards to generation. std::map obj_to_index; m->c_shared_object_data.nshared_first_page = toI(m->part6.size()); @@ -1441,8 +1383,7 @@ QPDF::calculateLinearizationData(std::map const& object_stream_data) stopOnError("shared object hint table has wrong number of entries"); } - // Now compute the list of shared objects for each page after the - // first page. + // Now compute the list of shared objects for each page after the first page. for (size_t i = 1; i < toS(npages); ++i) { CHPageOffsetEntry& pe = m->c_page_offset_data.entries.at(i); @@ -1520,20 +1461,17 @@ QPDF::outputLengthNextN( std::map const& lengths, std::map const& obj_renumber) { - // Figure out the length of a series of n consecutive objects in - // the output file starting with whatever object in_object from - // the input file mapped to. + // Figure out the length of a series of n consecutive objects in the output file starting with + // whatever object in_object from the input file mapped to. if (obj_renumber.count(in_object) == 0) { - stopOnError("found object that is not renumbered while" - " writing linearization data"); + stopOnError("found object that is not renumbered while writing linearization data"); } int first = (*(obj_renumber.find(in_object))).second; int length = 0; for (int i = 0; i < n; ++i) { if (lengths.count(first + i) == 0) { - stopOnError("found item with unknown length" - " while writing linearization data"); + stopOnError("found item with unknown length while writing linearization data"); } length += toI((*(lengths.find(first + toI(i)))).second); } @@ -1548,16 +1486,14 @@ QPDF::calculateHPageOffset( { // Page Offset Hint Table - // We are purposely leaving some values set to their initial zero - // values. + // We are purposely leaving some values set to their initial zero values. std::vector const& pages = getAllPages(); size_t npages = pages.size(); CHPageOffset& cph = m->c_page_offset_data; std::vector& cphe = cph.entries; - // Calculate minimum and maximum values for number of objects per - // page and page length. + // Calculate minimum and maximum values for number of objects per page and page length. int min_nobjects = cphe.at(0).nobjects; int max_nobjects = min_nobjects; @@ -1572,11 +1508,11 @@ QPDF::calculateHPageOffset( phe = std::vector(npages); for (unsigned int i = 0; i < npages; ++i) { - // Calculate values for each page, assigning full values to - // the delta items. They will be adjusted later. + // Calculate values for each page, assigning full values to the delta items. They will be + // adjusted later. - // Repeat calculations for page 0 so we can assign to phe[i] - // without duplicating those assignments. + // Repeat calculations for page 0 so we can assign to phe[i] without duplicating those + // assignments. int nobjects = cphe.at(i).nobjects; int length = outputLengthNextN(pages.at(i).getObjectID(), nobjects, lengths, obj_renumber); @@ -1604,11 +1540,10 @@ QPDF::calculateHPageOffset( ph.nbits_shared_identifier = nbits(m->c_shared_object_data.nshared_total); ph.shared_denominator = 4; // doesn't matter - // It isn't clear how to compute content offset and content - // length. Since we are not interleaving page objects with the - // content stream, we'll use the same values for content length as - // page length. We will use 0 as content offset because this is - // what Adobe does (implementation note 127) and pdlin as well. + // It isn't clear how to compute content offset and content length. Since we are not + // interleaving page objects with the content stream, we'll use the same values for content + // length as page length. We will use 0 as content offset because this is what Adobe does + // (implementation note 127) and pdlin as well. ph.nbits_delta_content_length = ph.nbits_delta_page_length; ph.min_content_length = ph.min_page_length; @@ -1616,8 +1551,8 @@ QPDF::calculateHPageOffset( // Adjust delta entries if ((phe.at(i).delta_nobjects < min_nobjects) || (phe.at(i).delta_page_length < min_length)) { - stopOnError("found too small delta nobjects or delta page length" - " while writing linearization data"); + stopOnError("found too small delta nobjects or delta page length while writing " + "linearization data"); } phe.at(i).delta_nobjects -= min_nobjects; phe.at(i).delta_page_length -= min_length; @@ -1669,8 +1604,7 @@ QPDF::calculateHSharedObject( for (size_t i = 0; i < toS(cso.nshared_total); ++i) { // Adjust deltas if (soe.at(i).delta_group_length < min_length) { - stopOnError("found too small group length while" - " writing linearization data"); + stopOnError("found too small group length while writing linearization data"); } soe.at(i).delta_group_length -= min_length; } @@ -1700,14 +1634,13 @@ template static void write_vector_int(BitWriter& w, int nitems, std::vector& vec, int bits, int_type T::*field) { - // nitems times, write bits bits from the given field of the ith - // vector to the given bit writer. + // nitems times, write bits bits from the given field of the ith vector to the given bit writer. for (size_t i = 0; i < QIntC::to_size(nitems); ++i) { w.writeBits(QIntC::to_ulonglong(vec.at(i).*field), QIntC::to_size(bits)); } - // The PDF spec says that each hint table starts at a byte - // boundary. Each "row" actually must start on a byte boundary. + // The PDF spec says that each hint table starts at a byte boundary. Each "row" actually must + // start on a byte boundary. w.flush(); } @@ -1721,8 +1654,8 @@ write_vector_vector( int bits, std::vector T::*vec2) { - // nitems1 times, write nitems2 (from the ith element of vec1) items - // from the vec2 vector field of the ith item of vec1. + // nitems1 times, write nitems2 (from the ith element of vec1) items from the vec2 vector field + // of the ith item of vec1. for (size_t i1 = 0; i1 < QIntC::to_size(nitems1); ++i1) { for (size_t i2 = 0; i2 < QIntC::to_size(vec1.at(i1).*nitems2); ++i2) { w.writeBits(QIntC::to_ulonglong((vec1.at(i1).*vec2).at(i2)), QIntC::to_size(bits)); @@ -1835,8 +1768,8 @@ QPDF::generateHintStream( calculateHSharedObject(xref, lengths, obj_renumber); calculateHOutline(xref, lengths, obj_renumber); - // Write the hint stream itself into a compressed memory buffer. - // Write through a counter so we can get offsets. + // Write the hint stream itself into a compressed memory buffer. Write through a counter so we + // can get offsets. Pl_Buffer hint_stream("hint stream"); Pl_Flate f("compress hint stream", &hint_stream, Pl_Flate::a_deflate); Pl_Count c("count", &f); diff --git a/libqpdf/QPDF_optimization.cc b/libqpdf/QPDF_optimization.cc index ab06f158..91da7564 100644 --- a/libqpdf/QPDF_optimization.cc +++ b/libqpdf/QPDF_optimization.cc @@ -64,9 +64,8 @@ QPDF::optimize( return; } - // The PDF specification indicates that /Outlines is supposed to - // be an indirect reference. Force it to be so if it exists and - // is direct. (This has been seen in the wild.) + // The PDF specification indicates that /Outlines is supposed to be an indirect reference. Force + // it to be so if it exists and is direct. (This has been seen in the wild.) QPDFObjectHandle root = getRoot(); if (root.getKey("/Outlines").isDictionary()) { QPDFObjectHandle outlines = root.getKey("/Outlines"); @@ -76,8 +75,8 @@ QPDF::optimize( } } - // Traverse pages tree pushing all inherited resources down to the - // page level. This also initializes m->all_pages. + // Traverse pages tree pushing all inherited resources down to the page level. This also + // initializes m->all_pages. pushInheritedAttributesToPage(allow_changes, false); // Traverse pages @@ -102,12 +101,10 @@ QPDF::optimize( } for (auto const& key: root.getKeys()) { - // Technically, /I keys from /Thread dictionaries are supposed - // to be handled separately, but we are going to disregard - // that specification for now. There is loads of evidence - // that pdlin and Acrobat both disregard things like this from - // time to time, so this is almost certain not to cause any - // problems. + // Technically, /I keys from /Thread dictionaries are supposed to be handled separately, but + // we are going to disregard that specification for now. There is loads of evidence that + // pdlin and Acrobat both disregard things like this from time to time, so this is almost + // certain not to cause any problems. updateObjectMaps( ObjUser(ObjUser::ou_root_key, key), root.getKey(key), skip_stream_parameters); } @@ -130,23 +127,20 @@ QPDF::pushInheritedAttributesToPage() void QPDF::pushInheritedAttributesToPage(bool allow_changes, bool warn_skipped_keys) { - // Traverse pages tree pushing all inherited resources down to the - // page level. + // Traverse pages tree pushing all inherited resources down to the page level. - // The record of whether we've done this is cleared by - // updateAllPagesCache(). If we're warning for skipped keys, - // re-traverse unconditionally. + // The record of whether we've done this is cleared by updateAllPagesCache(). If we're warning + // for skipped keys, re-traverse unconditionally. if (m->pushed_inherited_attributes_to_pages && (!warn_skipped_keys)) { return; } - // Calling getAllPages() resolves any duplicated page objects, - // repairs broken nodes, and detects loops, so we don't have to do - // those activities here. + // Calling getAllPages() resolves any duplicated page objects, repairs broken nodes, and detects + // loops, so we don't have to do those activities here. getAllPages(); - // key_ancestors is a mapping of page attribute keys to a stack of - // Pages nodes that contain values for them. + // key_ancestors is a mapping of page attribute keys to a stack of Pages nodes that contain + // values for them. std::map> key_ancestors; pushInheritedAttributesToPageInternal( m->trailer.getKey("/Root").getKey("/Pages"), @@ -168,10 +162,9 @@ QPDF::pushInheritedAttributesToPageInternal( bool allow_changes, bool warn_skipped_keys) { - // Make a list of inheritable keys. Only the keys /MediaBox, - // /CropBox, /Resources, and /Rotate are inheritable - // attributes. Push this object onto the stack of pages nodes - // that have values for this attribute. + // Make a list of inheritable keys. Only the keys /MediaBox, /CropBox, /Resources, and /Rotate + // are inheritable attributes. Push this object onto the stack of pages nodes that have values + // for this attribute. std::set inheritable_keys; for (auto const& key: cur_pages.getKeys()) { @@ -183,9 +176,7 @@ QPDF::pushInheritedAttributesToPageInternal( m->file->getName(), m->last_object_description, m->file->getLastOffset(), - "optimize detected an " - "inheritable attribute when called " - "in no-change mode"); + "optimize detected an inheritable attribute when called in no-change mode"); } // This is an inheritable resource @@ -194,9 +185,8 @@ QPDF::pushInheritedAttributesToPageInternal( QTC::TC("qpdf", "QPDF opt direct pages resource", oh.isIndirect() ? 0 : 1); if (!oh.isIndirect()) { if (!oh.isScalar()) { - // Replace shared direct object non-scalar - // resources with indirect objects to avoid - // copying large structures around. + // Replace shared direct object non-scalar resources with indirect objects to + // avoid copying large structures around. cur_pages.replaceKey(key, makeIndirectObject(oh)); oh = cur_pages.getKey(key); } else { @@ -208,14 +198,12 @@ QPDF::pushInheritedAttributesToPageInternal( if (key_ancestors[key].size() > 1) { QTC::TC("qpdf", "QPDF opt key ancestors depth > 1"); } - // Remove this resource from this node. It will be - // reattached at the page level. + // Remove this resource from this node. It will be reattached at the page level. cur_pages.removeKey(key); } else if (!((key == "/Type") || (key == "/Parent") || (key == "/Kids") || (key == "/Count"))) { - // Warn when flattening, but not if the key is at the top - // level (i.e. "/Parent" not set), as we don't change these; - // but flattening removes intermediate /Pages nodes. + // Warn when flattening, but not if the key is at the top level (i.e. "/Parent" not + // set), as we don't change these; but flattening removes intermediate /Pages nodes. if ((warn_skipped_keys) && (cur_pages.hasKey("/Parent"))) { QTC::TC("qpdf", "QPDF unknown key not inherited"); setLastObjectDescription("Pages object", cur_pages.getObjGen()); @@ -224,24 +212,21 @@ QPDF::pushInheritedAttributesToPageInternal( m->last_object_description, 0, ("Unknown key " + key + - " in /Pages object" - " is being discarded as a result of" - " flattening the /Pages tree")); + " in /Pages object is being discarded as a result of flattening the /Pages " + "tree")); } } } - // Process descendant nodes. This method does not perform loop - // detection because all code paths that lead here follow a call - // to getAllPages, which already throws an exception in the event + // Process descendant nodes. This method does not perform loop detection because all code paths + // that lead here follow a call to getAllPages, which already throws an exception in the event // of a loop in the pages tree. for (auto& kid: cur_pages.getKey("/Kids").aitems()) { if (kid.isDictionaryOfType("/Pages")) { pushInheritedAttributesToPageInternal( kid, key_ancestors, allow_changes, warn_skipped_keys); } else { - // Add all available inheritable attributes not present in - // this object to this object. + // Add all available inheritable attributes not present in this object to this object. for (auto const& iter: key_ancestors) { std::string const& key = iter.first; if (!kid.hasKey(key)) { @@ -254,10 +239,9 @@ QPDF::pushInheritedAttributesToPageInternal( } } - // For each inheritable key, pop the stack. If the stack - // becomes empty, remove it from the map. That way, the - // invariant that the list of keys in key_ancestors is exactly - // those keys for which inheritable attributes are available. + // For each inheritable key, pop the stack. If the stack becomes empty, remove it from the map. + // That way, the invariant that the list of keys in key_ancestors is exactly those keys for + // which inheritable attributes are available. if (!inheritable_keys.empty()) { QTC::TC("qpdf", "QPDF opt inheritable keys"); @@ -291,8 +275,7 @@ QPDF::updateObjectMapsInternal( QPDFObjGen::set& visited, bool top) { - // Traverse the object tree from this point taking care to avoid - // crossing page boundaries. + // Traverse the object tree from this point taking care to avoid crossing page boundaries. bool is_page_node = false; @@ -332,8 +315,7 @@ QPDF::updateObjectMapsInternal( for (auto const& key: dict.getKeys()) { if (is_page_node && (key == "/Thumb")) { - // Traverse page thumbnail dictionaries as a special - // case. + // Traverse page thumbnail dictionaries as a special case. updateObjectMapsInternal( ObjUser(ObjUser::ou_thumb, ou.pageno), dict.getKey(key), @@ -345,8 +327,7 @@ QPDF::updateObjectMapsInternal( } else if ( ((ssp >= 1) && (key == "/Length")) || ((ssp >= 2) && ((key == "/Filter") || (key == "/DecodeParms")))) { - // Don't traverse into stream parameters that we are - // not going to write. + // Don't traverse into stream parameters that we are not going to write. } else { updateObjectMapsInternal( ou, dict.getKey(key), skip_stream_parameters, visited, false); @@ -362,9 +343,8 @@ QPDF::filterCompressedObjects(std::map const& object_stream_data) return; } - // Transform object_to_obj_users and obj_user_to_objects so that - // they refer only to uncompressed objects. If something is a - // user of a compressed object, then it is really a user of the + // Transform object_to_obj_users and obj_user_to_objects so that they refer only to uncompressed + // objects. If something is a user of a compressed object, then it is really a user of the // object stream that contains it. std::map> t_obj_user_to_objects; diff --git a/libqpdf/QPDF_pages.cc b/libqpdf/QPDF_pages.cc index 81fd11a3..e03dabc8 100644 --- a/libqpdf/QPDF_pages.cc +++ b/libqpdf/QPDF_pages.cc @@ -4,55 +4,42 @@ #include #include -// In support of page manipulation APIs, these methods internally -// maintain state about pages in a pair of data structures: all_pages, -// which is a vector of page objects, and pageobj_to_pages_pos, which -// maps a page object to its position in the all_pages array. -// Unfortunately, the getAllPages() method returns a const reference -// to all_pages and has been in the public API long before the -// introduction of mutation APIs, so we're pretty much stuck with it. -// Anyway, there are lots of calls to it in the library, so the -// efficiency of having it cached is probably worth keeping it. At one -// point, I had partially implemented a helper class specifically for -// the pages tree, but once you work in all the logic that handles -// repairing the /Type keys of page tree nodes (both /Pages and /Page) -// and deal with duplicate pages, it's just as complex and less -// efficient than what's here. So, in spite of the fact that a const -// reference is returned, the current code is fine and does not need -// to be replaced. A partial implementation of QPDFPagesTree is in -// github in attic in case there is ever a reason to resurrect it. -// There are additional notes in README-maintainer, which also refers -// to this comment. +// In support of page manipulation APIs, these methods internally maintain state about pages in a +// pair of data structures: all_pages, which is a vector of page objects, and pageobj_to_pages_pos, +// which maps a page object to its position in the all_pages array. Unfortunately, the getAllPages() +// method returns a const reference to all_pages and has been in the public API long before the +// introduction of mutation APIs, so we're pretty much stuck with it. Anyway, there are lots of +// calls to it in the library, so the efficiency of having it cached is probably worth keeping it. +// At one point, I had partially implemented a helper class specifically for the pages tree, but +// once you work in all the logic that handles repairing the /Type keys of page tree nodes (both +// /Pages and /Page) and deal with duplicate pages, it's just as complex and less efficient than +// what's here. So, in spite of the fact that a const reference is returned, the current code is +// fine and does not need to be replaced. A partial implementation of QPDFPagesTree is in github in +// attic in case there is ever a reason to resurrect it. There are additional notes in +// README-maintainer, which also refers to this comment. -// The goal of this code is to ensure that the all_pages vector, which -// users may have a reference to, and the pageobj_to_pages_pos map, -// which users will not have access to, remain consistent outside of -// any call to the library. As long as users only touch the /Pages -// structure through page-specific API calls, they never have to worry -// about anything, and this will also stay consistent. If a user -// touches anything about the /Pages structure outside of these calls -// (such as by directly looking up and manipulating the underlying -// objects), they can call updatePagesCache() to bring things back in -// sync. +// The goal of this code is to ensure that the all_pages vector, which users may have a reference +// to, and the pageobj_to_pages_pos map, which users will not have access to, remain consistent +// outside of any call to the library. As long as users only touch the /Pages structure through +// page-specific API calls, they never have to worry about anything, and this will also stay +// consistent. If a user touches anything about the /Pages structure outside of these calls (such +// as by directly looking up and manipulating the underlying objects), they can call +// updatePagesCache() to bring things back in sync. -// If the user doesn't ever use the page manipulation APIs, then qpdf -// leaves the /Pages structure alone. If the user does use the APIs, -// then we push all inheritable objects down and flatten the /Pages -// tree. This makes it easier for us to keep /Pages, all_pages, and -// pageobj_to_pages_pos internally consistent at all times. +// If the user doesn't ever use the page manipulation APIs, then qpdf leaves the /Pages structure +// alone. If the user does use the APIs, then we push all inheritable objects down and flatten the +// /Pages tree. This makes it easier for us to keep /Pages, all_pages, and pageobj_to_pages_pos +// internally consistent at all times. -// Responsibility for keeping all_pages, pageobj_to_pages_pos, and the -// Pages structure consistent should remain in as few places as -// possible. As of initial writing, only flattenPagesTree, -// insertPage, and removePage, along with methods they call, are -// concerned with it. Everything else goes through one of those -// methods. +// Responsibility for keeping all_pages, pageobj_to_pages_pos, and the Pages structure consistent +// should remain in as few places as possible. As of initial writing, only flattenPagesTree, +// insertPage, and removePage, along with methods they call, are concerned with it. Everything else +// goes through one of those methods. std::vector const& QPDF::getAllPages() { - // Note that pushInheritedAttributesToPage may also be used to - // initialize m->all_pages. + // Note that pushInheritedAttributesToPage may also be used to initialize m->all_pages. if (m->all_pages.empty()) { m->ever_called_get_all_pages = true; QPDFObjGen::set visited; @@ -65,9 +52,8 @@ QPDF::getAllPages() // loop -- will be detected again and reported later break; } - // Files have been found in the wild where /Pages in the - // catalog points to the first page. Try to work around - // this and similar cases with this heuristic. + // Files have been found in the wild where /Pages in the catalog points to the first + // page. Try to work around this and similar cases with this heuristic. if (!warned) { getRoot().warnIfPossible("document page tree root (root -> /Pages) doesn't point" " to the root of the page tree; attempting to correct"); @@ -118,8 +104,8 @@ QPDF::getAllPagesInternal( kid = makeIndirectObject(kid); kids.setArrayItem(i, kid); } else if (!seen.add(kid)) { - // Make a copy of the page. This does the same as - // shallowCopyPage in QPDFPageObjectHelper. + // Make a copy of the page. This does the same as shallowCopyPage in + // QPDFPageObjectHelper. QTC::TC("qpdf", "QPDF resolve duplicated page object"); cur_node.warnIfPossible( "kid " + std::to_string(i) + @@ -141,9 +127,8 @@ QPDF::getAllPagesInternal( void QPDF::updateAllPagesCache() { - // Force regeneration of the pages cache. We force immediate - // recalculation of all_pages since users may have references to - // it that they got from calls to getAllPages(). We can defer + // Force regeneration of the pages cache. We force immediate recalculation of all_pages since + // users may have references to it that they got from calls to getAllPages(). We can defer // recalculation of pageobj_to_pages_pos until needed. QTC::TC("qpdf", "QPDF updateAllPagesCache"); m->all_pages.clear(); @@ -155,25 +140,23 @@ QPDF::updateAllPagesCache() void QPDF::flattenPagesTree() { - // If not already done, flatten the /Pages structure and - // initialize pageobj_to_pages_pos. + // If not already done, flatten the /Pages structure and initialize pageobj_to_pages_pos. if (!m->pageobj_to_pages_pos.empty()) { return; } - // Push inherited objects down to the /Page level. As a side - // effect m->all_pages will also be generated. + // Push inherited objects down to the /Page level. As a side effect m->all_pages will also be + // generated. pushInheritedAttributesToPage(true, true); QPDFObjectHandle pages = getRoot().getKey("/Pages"); size_t const len = m->all_pages.size(); for (size_t pos = 0; pos < len; ++pos) { - // Populate pageobj_to_pages_pos and fix parent pointer. There - // should be no duplicates at this point because - // pushInheritedAttributesToPage calls getAllPages which - // resolves duplicates. + // Populate pageobj_to_pages_pos and fix parent pointer. There should be no duplicates at + // this point because pushInheritedAttributesToPage calls getAllPages which resolves + // duplicates. insertPageobjToPage(m->all_pages.at(pos), toI(pos), true); m->all_pages.at(pos).replaceKey("/Parent", pages); } @@ -191,16 +174,14 @@ QPDF::insertPageobjToPage(QPDFObjectHandle const& obj, int pos, bool check_dupli QPDFObjGen og(obj.getObjGen()); if (check_duplicate) { if (!m->pageobj_to_pages_pos.insert(std::make_pair(og, pos)).second) { - // The library never calls insertPageobjToPage in a way - // that causes this to happen. + // The library never calls insertPageobjToPage in a way that causes this to happen. setLastObjectDescription("page " + std::to_string(pos) + " (numbered from zero)", og); throw QPDFExc( qpdf_e_pages, m->file->getName(), m->last_object_description, 0, - "duplicate page reference found;" - " this would cause loss of data"); + "duplicate page reference found; this would cause loss of data"); } } else { m->pageobj_to_pages_pos[og] = pos; @@ -210,8 +191,7 @@ QPDF::insertPageobjToPage(QPDFObjectHandle const& obj, int pos, bool check_dupli void QPDF::insertPage(QPDFObjectHandle newpage, int pos) { - // pos is numbered from 0, so pos = 0 inserts at the beginning and - // pos = npages adds to the end. + // pos is numbered from 0, so pos = 0 inserts at the beginning and pos = npages adds to the end. flattenPagesTree(); @@ -233,10 +213,9 @@ QPDF::insertPage(QPDFObjectHandle newpage, int pos) QTC::TC( "qpdf", "QPDF insert page", - (pos == 0) ? 0 : // insert at beginning - (pos == toI(m->all_pages.size())) ? 1 - : // at end - 2); // insert in middle + (pos == 0) ? 0 : // insert at beginning + (pos == toI(m->all_pages.size())) ? 1 // at end + : 2); // insert in middle auto og = newpage.getObjGen(); if (m->pageobj_to_pages_pos.count(og)) { @@ -265,10 +244,9 @@ QPDF::removePage(QPDFObjectHandle page) QTC::TC( "qpdf", "QPDF remove page", - (pos == 0) ? 0 : // remove at beginning - (pos == toI(m->all_pages.size() - 1)) ? 1 - : // end - 2); // remove in middle + (pos == 0) ? 0 : // remove at beginning + (pos == toI(m->all_pages.size() - 1)) ? 1 // end + : 2); // remove in middle QPDFObjectHandle pages = getRoot().getKey("/Pages"); QPDFObjectHandle kids = pages.getKey("/Kids"); -- cgit v1.2.3-54-g00ecf