Precheck streams by default

There is no need for a --precheck-streams option. We can do the
precheck without imposing any penalty, only re-encoding the stream if
it fails the first time.

8 files changed, 32 insertions, 81 deletions

diff --git a/include/qpdf/QPDFWriter.hh b/include/qpdf/QPDFWriter.hh
index c4bc7846..2f878492 100644
--- a/include/qpdf/QPDFWriter.hh
+++ b/include/qpdf/QPDFWriter.hh
@@ -200,17 +200,6 @@ class QPDFWriter
     QPDF_DLL
     void setQDFMode(bool);
 
-    // Enable stream precheck mode. In this mode, all filterable
-    // streams are checked by actually attempting to decode them
-    // before filtering. This may add significant time to the process
-    // of writing the data because all streams from the input must be
-    // read twice, but it enables the raw stream data to be preserved
-    // even in cases where qpdf would run into errors decoding the
-    // stream after it determines that it should be able to do it.
-    // Examples would include compressed data with errors in it.
-    QPDF_DLL
-    void setPrecheckStreams(bool);
-
     // Preserve unreferenced objects. The default behavior is to
     // discard any object that is not visited during a traversal of
     // the object structure from the trailer.
@@ -495,7 +484,6 @@ class QPDFWriter
     qpdf_stream_decode_level_e stream_decode_level;
     bool stream_decode_level_set;
     bool qdf_mode;
-    bool precheck_streams;
     bool preserve_unreferenced_objects;
     bool newline_before_endstream;
     bool static_id;
diff --git a/libqpdf/QPDFWriter.cc b/libqpdf/QPDFWriter.cc
index fe25853a..06945c49 100644
--- a/libqpdf/QPDFWriter.cc
+++ b/libqpdf/QPDFWriter.cc
@@ -59,7 +59,6 @@ QPDFWriter::init()
     stream_decode_level = qpdf_dl_none;
     stream_decode_level_set = false;
     qdf_mode = false;
-    precheck_streams = false;
     preserve_unreferenced_objects = false;
     newline_before_endstream = false;
     static_id = false;
@@ -216,12 +215,6 @@ QPDFWriter::setQDFMode(bool val)
 }
 
 void
-QPDFWriter::setPrecheckStreams(bool val)
-{
-    this->precheck_streams = val;
-}
-
-void
 QPDFWriter::setPreserveUnreferencedObjects(bool val)
 {
     this->preserve_unreferenced_objects = val;
@@ -1590,34 +1583,32 @@ QPDFWriter::unparseObject(QPDFObjectHandle object, int level,
 
 	flags |= f_stream;
 
-        if (filter && this->precheck_streams)
+        PointerHolder<Buffer> stream_data;
+        bool filtered = false;
+        for (int attempt = 1; attempt <= 2; ++attempt)
         {
-            try
+            pushPipeline(new Pl_Buffer("stream data"));
+            activatePipelineStack();
+
+            filtered =
+                object.pipeStreamData(
+                    this->pipeline,
+                    (((filter && normalize) ? qpdf_ef_normalize : 0) |
+                     ((filter && compress) ? qpdf_ef_compress : 0)),
+                    (filter
+                     ? (uncompress ? qpdf_dl_all : this->stream_decode_level)
+                     : qpdf_dl_none));
+            popPipelineStack(&stream_data);
+            if (filter && (! filtered))
             {
-                QTC::TC("qpdf", "QPDFWriter precheck stream");
-                Pl_Discard discard;
-                filter = object.pipeStreamData(
-                    &discard, 0, qpdf_dl_all, true);
+                // Try again
+                filter = false;
             }
-            catch (std::exception&)
+            else
             {
-                filter = false;
+                break;
             }
         }
-
-	pushPipeline(new Pl_Buffer("stream data"));
-	activatePipelineStack();
-
-	bool filtered =
-	    object.pipeStreamData(
-                this->pipeline,
-                (((filter && normalize) ? qpdf_ef_normalize : 0) |
-                 ((filter && compress) ? qpdf_ef_compress : 0)),
-                (filter
-                 ? (uncompress ? qpdf_dl_all : this->stream_decode_level)
-                 : qpdf_dl_none));
-	PointerHolder<Buffer> stream_data;
-	popPipelineStack(&stream_data);
 	if (filtered)
 	{
 	    flags |= f_filtered;
diff --git a/manual/qpdf-manual.xml b/manual/qpdf-manual.xml
index 14e8b8bc..0e7d05f7 100644
--- a/manual/qpdf-manual.xml
+++ b/manual/qpdf-manual.xml
@@ -997,23 +997,6 @@ outfile.pdf</option>
       </listitem>
      </varlistentry>
      <varlistentry>
-      <term><option>--precheck-streams</option></term>
-      <listitem>
-       <para>
-        Tells qpdf to precheck each stream for the ability to decode
-        it. Ordinarily qpdf tries to decode streams that it thinks it
-        can decode based on the filters, and if there ends up being an
-        error when actually trying to do the decode, the stream data
-        is truncated. This flag causes qpdf to actually read the
-        stream fully before deciding whether to filter the stream.
-        This option will slow qpdf down since it will have to read the
-        stream twice, but it allows raw stream data to be preserved in
-        cases where the decoding of the stream would fail for some
-        reason. This may be useful in working with some damaged files.
-       </para>
-      </listitem>
-     </varlistentry>
-     <varlistentry>
       <term><option>--preserve-unreferenced</option></term>
       <listitem>
        <para>
diff --git a/qpdf/qpdf.cc b/qpdf/qpdf.cc
index df2ba225..ade0d4f2 100644
--- a/qpdf/qpdf.cc
+++ b/qpdf/qpdf.cc
@@ -87,7 +87,6 @@ struct Options
         object_stream_mode(qpdf_o_preserve),
         ignore_xref_streams(false),
         qdf_mode(false),
-        precheck_streams(false),
         preserve_unreferenced_objects(false),
         newline_before_endstream(false),
         show_npages(false),
@@ -149,7 +148,6 @@ struct Options
     qpdf_object_stream_e object_stream_mode;
     bool ignore_xref_streams;
     bool qdf_mode;
-    bool precheck_streams;
     bool preserve_unreferenced_objects;
     bool newline_before_endstream;
     std::string min_version;
@@ -371,7 +369,6 @@ familiar with the PDF file format or who are PDF developers.\n\
 --suppress-recovery       prevents qpdf from attempting to recover damaged files\n\
 --object-streams=mode     controls handing of object streams\n\
 --ignore-xref-streams     tells qpdf to ignore any cross-reference streams\n\
---precheck-streams        precheck ability to decode streams\n\
 --preserve-unreferenced   preserve unreferenced objects\n\
 --newline-before-endstream  always put a newline before endstream\n\
 --qdf                     turns on \"QDF mode\" (below)\n\
@@ -1467,10 +1464,6 @@ static void parse_options(int argc, char* argv[], Options& o)
             {
                 o.qdf_mode = true;
             }
-            else if (strcmp(arg, "precheck-streams") == 0)
-            {
-                o.precheck_streams = true;
-            }
             else if (strcmp(arg, "preserve-unreferenced") == 0)
             {
                 o.preserve_unreferenced_objects = true;
@@ -2094,10 +2087,6 @@ static void set_writer_options(QPDF& pdf, Options& o, QPDFWriter& w)
     {
         w.setQDFMode(true);
     }
-    if (o.precheck_streams)
-    {
-        w.setPrecheckStreams(true);
-    }
     if (o.preserve_unreferenced_objects)
     {
         w.setPreserveUnreferencedObjects(true);
diff --git a/qpdf/qpdf.testcov b/qpdf/qpdf.testcov
index 2a157c91..c1355d59 100644
--- a/qpdf/qpdf.testcov
+++ b/qpdf/qpdf.testcov
@@ -277,7 +277,6 @@ QPDFObjectHandle treat word as string 0
 QPDFObjectHandle found fake 1
 QPDFObjectHandle no val for last key 0
 QPDF resolve failure to null 0
-QPDFWriter precheck stream 0
 QPDFWriter preserve unreferenced standard 0
 QPDFObjectHandle non-stream in parsecontent 0
 QPDFObjectHandle errors in parsecontent 0
diff --git a/qpdf/qtest/qpdf.test b/qpdf/qtest/qpdf.test
index 97d73277..cf8c88ca 100644
--- a/qpdf/qtest/qpdf.test
+++ b/qpdf/qtest/qpdf.test
@@ -918,27 +918,20 @@ $td->runtest("check output",
 show_ntests();
 # ----------
 $td->notify("--- Precheck streams ---");
-$n_tests += 4;
+$n_tests += 2;
 
-$td->runtest("bad stream without precheck",
+$td->runtest("bad stream",
 	     {$td->COMMAND => "qpdf --static-id bad-data.pdf a.pdf"},
 	     {$td->FILE => "bad-data.out", $td->EXIT_STATUS => 3},
 	     $td->NORMALIZE_NEWLINES);
 $td->runtest("check output",
 	     {$td->FILE => "a.pdf"},
 	     {$td->FILE => "bad-data-out.pdf"});
-$td->runtest("bad stream with precheck",
-	     {$td->COMMAND =>
-                  "qpdf --static-id --precheck-streams bad-data.pdf a.pdf"},
-	     {$td->STRING => "", $td->EXIT_STATUS => 0},
-	     $td->NORMALIZE_NEWLINES);
-$td->runtest("check output",
-	     {$td->FILE => "a.pdf"},
-	     {$td->FILE => "bad-data-precheck.pdf"});
+
 show_ntests();
 # ----------
 $td->notify("--- Decode levels ---");
-$n_tests += 10;
+$n_tests += 12;
 
 # image-streams.pdf is the output of examples/pdf-create.
 # examples/pdf-create validates the actual image data.
@@ -962,6 +955,14 @@ $td->runtest("check finds bad jpeg data",
              {$td->FILE => "bad-jpeg-check.out",
               $td->EXIT_STATUS => 3},
              $td->NORMALIZE_NEWLINES);
+$td->runtest("precheck detects bad jpeg data",
+             {$td->COMMAND => "qpdf --static-id --decode-level=all" .
+                  " bad-jpeg.pdf a.pdf"},
+             {$td->FILE => "bad-jpeg.out", $td->EXIT_STATUS => 3},
+             $td->NORMALIZE_NEWLINES);
+$td->runtest("check file",
+             {$td->FILE => "a.pdf"},
+             {$td->FILE => "bad-jpeg-out.pdf"});
 $td->runtest("get data",
              {$td->COMMAND => "qpdf --show-object=6" .
                   " --filtered-stream-data bad-jpeg.pdf"},
diff --git a/qpdf/qtest/qpdf/bad-data-out.pdf b/qpdf/qtest/qpdf/bad-data-out.pdf
index f4300662..4314025a 100644
--- a/qpdf/qtest/qpdf/bad-data-out.pdf
+++ b/qpdf/qtest/qpdf/bad-data-out.pdf
diff --git a/qpdf/qtest/qpdf/bad-data-precheck.pdf b/qpdf/qtest/qpdf/bad-data-precheck.pdf
deleted file mode 100644
index 4314025a..00000000
--- a/qpdf/qtest/qpdf/bad-data-precheck.pdf
+++ /dev/null