Have qpdf-test-compare handle certain /ID differences

Rationale is in comments in the code. Deterministic ID tests explicitly do not use this tool.

Have qpdf-test-compare handle certain /ID differences
Rationale is in comments in the code. Deterministic ID tests explicitly do not use this tool.
Jay Berkenbilt
1 parent a238b4b0
Showing 6 changed files with 37 additions and 3 deletions
compare-for-test/qpdf-test-compare.cc
compare-for-test/qtest/compare.test
compare-for-test/qtest/compare/diff-id.pdf
compare-for-test/qtest/compare/zlib-new-id.pdf
compare-for-test/qtest/compare/zlib-new-id1.pdf
compare-for-test/qtest/compare/zlib-new-id2.pdf
@@ -24,7 +24,7 @@ usage()
               << std::endl
               << "If the files match, the output is the expected file. Otherwise, it is"
               << std::endl
-              << "the actual file. Read comments in the test suite for rationale." << std::endl;
+              << "the actual file. Read comments in the code for rationale." << std::endl;
     exit(2);
 }
  
@@ -50,6 +50,8 @@ compareObjects(std::string const&amp; label, QPDFObjectHandle act, QPDFObjectHandle 
         return label + ": different types";
     }
     if (act.isStream()) {
+        // Disregard stream lengths. The length of stream data is compared later, and we don't care
+        // about the length of compressed data as long as the uncompressed data matches.
         auto act_dict = act.getDict();
         auto exp_dict = exp.getDict();
         act_dict.removeKey("/Length");
@@ -59,6 +61,8 @@ compareObjects(std::string const&amp; label, QPDFObjectHandle act, QPDFObjectHandle 
             return label + ": stream dictionaries differ";
         }
         if (act_dict.getKey("/Type").isNameAndEquals("/XRef")) {
+            // Cross-reference streams will generally not match, but we have numerous tests that
+            // meaningfully ensure that xref streams are correct.
             QTC::TC("compare", "ignore data for xref stream");
             return "";
         }
@@ -103,6 +107,27 @@ compareObjects(std::string const&amp; label, QPDFObjectHandle act, QPDFObjectHandle 
     return "";
 }
  
+void
+cleanTrailer(QPDFObjectHandle& trailer)
+{
+    // If the trailer is an object stream, it will have /Length.
+    trailer.removeKey("/Length");
+    // Disregard the second half of /ID. This doesn't have anything directly to do with zlib, but
+    // lots of tests use --deterministic-id, and that is affected. The deterministic ID tests
+    // meaningfully exercise that deterministic IDs behave as expected, so for the rest of the
+    // tests, it's okay to ignore /ID[1]. If the two halves of /ID are the same, ignore both since
+    // this means qpdf completely generated the /ID rather than preserving the first half.
+    auto id = trailer.getKey("/ID");
+    if (id.isArray() && id.getArrayNItems() == 2) {
+        auto id0 = id.getArrayItem(0).unparse();
+        auto id1 = id.getArrayItem(1).unparse();
+        id.setArrayItem(1, "()"_qpdf);
+        if (id0 == id1) {
+            id.setArrayItem(0, "()"_qpdf);
+        }
+    }
+}
+
 std::string
 compare(char const* actual_filename, char const* expected_filename)
 {
@@ -118,8 +143,8 @@ compare(char const* actual_filename, char const* expected_filename)
  
     auto act_trailer = actual.getTrailer();
     auto exp_trailer = expected.getTrailer();
-    act_trailer.removeKey("/Length");
-    exp_trailer.removeKey("/Length");
+    cleanTrailer(act_trailer);
+    cleanTrailer(exp_trailer);
     auto trailer_diff = compareObjects("trailer", act_trailer, exp_trailer);
     if (!trailer_diff.empty()) {
         QTC::TC("compare", "different trailer");
@@ -52,6 +52,7 @@ my @diff = (
     ["diff-data-unc.pdf", "5,0: stream data differs"],
     ["diff-stream-dict.pdf", "4,0: stream dictionaries differ"],
     ["diff-object-type.pdf", "6,0: different types"],
+    ["diff-id.pdf", "trailer: object contents differ"],
     );
 $n_tests += 2 * scalar(@diff);
  
@@ -90,4 +91,12 @@ $td-&gt;runtest(&quot;compare object stream files (same)&quot;,
              {$td->COMMAND => "env QPDF_COMPARE_WHY=1 qpdf-test-compare ostream1.pdf ostream2.pdf"},
              {$td->FILE => "ostream2.pdf", $td->EXIT_STATUS => 0});
  
+$n_tests += 2;
+$td->runtest("files identical except /ID[1]",
+             {$td->COMMAND => "env QPDF_COMPARE_WHY=1 qpdf-test-compare zlib.pdf zlib-new-id.pdf"},
+             {$td->FILE => "zlib-new-id.pdf", $td->EXIT_STATUS => 0});
+$td->runtest("/ID[0] = /ID[1]",
+             {$td->COMMAND => "env QPDF_COMPARE_WHY=1 qpdf-test-compare zlib-new-id1.pdf zlib-new-id2.pdf"},
+             {$td->FILE => "zlib-new-id2.pdf", $td->EXIT_STATUS => 0});
+
 $td->report($n_tests);