aboutsummaryrefslogtreecommitdiffstats
path: root/libqpdf/InputSource.cc
diff options
context:
space:
mode:
authorJay Berkenbilt <ejb@ql.org>2017-08-06 14:42:01 +0200
committerJay Berkenbilt <ejb@ql.org>2017-08-11 03:30:32 +0200
commit296b679d6e3217cc112b7ed19b363b82356615ef (patch)
tree7cb9b3aa95c00da45cf1a7cf67a020c1ee54a6c9 /libqpdf/InputSource.cc
parentef8ae5449dc30782451beba64fdd0af86e1cb931 (diff)
downloadqpdf-296b679d6e3217cc112b7ed19b363b82356615ef.tar.zst
Implement findFirst and findLast in InputSource
Preparing to refactor some pattern searching code to use these instead of their own memchr loops. This should simplify the code that replaces PCRE.
Diffstat (limited to 'libqpdf/InputSource.cc')
-rw-r--r--libqpdf/InputSource.cc167
1 files changed, 167 insertions, 0 deletions
diff --git a/libqpdf/InputSource.cc b/libqpdf/InputSource.cc
index 79c889bf..69cafeb8 100644
--- a/libqpdf/InputSource.cc
+++ b/libqpdf/InputSource.cc
@@ -1,7 +1,10 @@
#include <qpdf/InputSource.hh>
#include <string.h>
+#include <stdexcept>
+#include <qpdf/QTC.hh>
#include <qpdf/PointerHolder.hh>
+
void
InputSource::setLastOffset(qpdf_offset_t offset)
{
@@ -39,3 +42,167 @@ InputSource::readLine(size_t max_line_length)
}
return std::string(buf);
}
+
+bool
+InputSource::findFirst(char const* start_chars,
+ qpdf_offset_t offset, size_t len,
+ Finder& finder)
+{
+ // Basic approach: search for the first character of start_chars
+ // starting from offset but not going past len (if len != 0). Once
+ // the first character is found, see if it is the beginning of a
+ // sequence of characters matching start_chars. If so, call
+ // finder.check() to do caller-specific additional checks. If not,
+ // keep searching.
+
+ // This code is tricky and highly subject to off-by-one or other
+ // edge case logic errors. See comments throughout that explain
+ // how we're not missing any edge cases. There are also tests
+ // specifically constructed to make sure we caught the edge cases
+ // in testing.
+
+ char buf[1025]; // size known to input_source.cc in libtests
+ // To enable us to guarantee null-termination, save an extra byte
+ // so that buf[size] is valid memory.
+ size_t size = sizeof(buf) - 1;
+ if ((strlen(start_chars) < 1) || (strlen(start_chars) > size))
+ {
+ throw std::logic_error(
+ "InputSource::findSource called with"
+ " too small or too large of a character sequence");
+ }
+
+ char* p = 0;
+ qpdf_offset_t buf_offset = offset;
+ size_t bytes_read = 0;
+
+ // Guarantee that we return from this loop. Each time through, we
+ // either return, advance p, or restart the loop with a condition
+ // that will cause return on the next pass. Eventually we will
+ // either be out of range or hit EOF, either of which forces us to
+ // return.
+ while (true)
+ {
+ // Do we need to read more data? Pretend size = 5, buf starts
+ // at 0, and start_chars has 3 characters. buf[5] is valid and
+ // null. If p == 2, start_chars could be buf[2] through
+ // buf[4], so p + strlen(start_chars) == buf + size is okay.
+ // If p points to buf[size], since strlen(start_chars) is
+ // always >= 1, this overflow test will be correct for that
+ // case regardless of start_chars.
+ if ((p == 0) || ((p + strlen(start_chars)) > (buf + bytes_read)))
+ {
+ if (p)
+ {
+ QTC::TC("libtests", "InputSource read next block",
+ ((p == buf + bytes_read) ? 0 : 1));
+ buf_offset += (p - buf);
+ }
+ this->seek(buf_offset, SEEK_SET);
+ // Read into buffer and zero out the rest of the buffer
+ // including buf[size]. We allocated an extra byte so that
+ // we could guarantee null termination as an extra
+ // protection against overrun when using string functions.
+ bytes_read = this->read(buf, size);
+ if (bytes_read < strlen(start_chars))
+ {
+ QTC::TC("libtests", "InputSource find EOF",
+ bytes_read == 0 ? 0 : 1);
+ return false;
+ }
+ memset(buf + bytes_read, '\0', 1 + (size - bytes_read));
+ p = buf;
+ }
+
+ // Search for the first character.
+ if ((p = static_cast<char*>(
+ memchr(p, start_chars[0], bytes_read - (p - buf)))) != 0)
+ {
+ if (p == buf)
+ {
+ QTC::TC("libtests", "InputSource found match at buf[0]");
+ }
+ // Found first letter.
+ if (len != 0)
+ {
+ // Make sure it's in range.
+ size_t p_relative_offset = (p - buf) + (buf_offset - offset);
+ if (p_relative_offset >= len)
+ {
+ // out of range
+ QTC::TC("libtests", "InputSource out of range");
+ return false;
+ }
+ }
+ if ((p + strlen(start_chars)) > (buf + bytes_read))
+ {
+ // If there are not enough bytes left in the file for
+ // start_chars, we will detect this on the next pass
+ // as EOF and return.
+ QTC::TC("libtests", "InputSource not enough bytes");
+ continue;
+ }
+
+ // See if p points to a sequence matching start_chars. We
+ // already checked above to make sure we are not going to
+ // overrun memory.
+ if (strncmp(p, start_chars, strlen(start_chars)) == 0)
+ {
+ // Call finder.check() with the input source
+ // positioned to the point of the match.
+ this->seek(buf_offset + (p - buf), SEEK_SET);
+ if (finder.check())
+ {
+ return true;
+ }
+ else
+ {
+ QTC::TC("libtests", "InputSource start_chars matched but not check");
+ }
+ }
+ else
+ {
+ QTC::TC("libtests", "InputSource first char matched but not string");
+ }
+ // This occurrence of the first character wasn't a match.
+ // Skip over it and keep searching.
+ ++p;
+ }
+ else
+ {
+ // Trigger reading the next block
+ p = buf + bytes_read;
+ }
+ }
+ throw std::logic_error("InputSource after while (true)");
+}
+
+bool
+InputSource::findLast(char const* start_chars,
+ qpdf_offset_t offset, size_t len,
+ Finder& finder)
+{
+ bool found = false;
+ qpdf_offset_t after_found_offset = 0;
+ qpdf_offset_t cur_offset = offset;
+ size_t cur_len = len;
+ while (this->findFirst(start_chars, cur_offset, cur_len, finder))
+ {
+ if (found)
+ {
+ QTC::TC("libtests", "InputSource findLast found more than one");
+ }
+ else
+ {
+ found = true;
+ }
+ after_found_offset = this->tell();
+ cur_offset = after_found_offset;
+ cur_len = len - (cur_offset - offset);
+ }
+ if (found)
+ {
+ this->seek(after_found_offset, SEEK_SET);
+ }
+ return found;
+}