Debian Codegraph @

podofo/test/unit/ContentStreamReaderTest.cpp

raw
1 // SPDX-FileCopyrightText: 2026 PoDoFo contributors
2 // SPDX-License-Identifier: MIT-0
3
4 #include <PdfTest.h>
5
6 using namespace std;
7 using namespace PoDoFo;
8
9 TEST_CASE("TestContentStreamReaderTerminates")
10 {
11 // Regression test for infinite loop in PdfContentStreamReader.
12 //
13 // Input "+\n\tH" (and similar short inputs like "-\0TT", ".\n\t\xf6",
14 // ".///") causes an infinite loop when fed into the content stream reader.
15 //
16 // Root cause: PdfTokenizer::DetermineDataType classifies a bare "+"
17 // (or "-", ".") as a candidate Number/Real, but std::from_chars fails
18 // to parse it. The recovery path re-enqueues the same token via
19 // EnqueueToken, then returns Unknown. On the next call to TryReadNextToken
20 // the enqueued token is dequeued, DetermineDataType fails again and
21 // re-enqueues it — creating an infinite loop.
22 //
23 // Inputs are taken from libFuzzer timeout artifacts (4 bytes each).
24 const string_view inputs[] = {
25 "+\n\tH"sv,
26 { "-\x00TT", 4 },
27 ".\n\t\xf6"sv,
28 ".///"sv,
29 { "\x00-(\xff", 4 },
30 };
31
32 for (auto& input : inputs)
33 {
34 INFO("Input bytes: " << input.size());
35
36 auto device = std::make_shared<SpanStreamDevice>(input);
37 PdfContentReaderArgs args;
38 args.Flags = PdfContentReaderFlags::SkipFollowFormXObjects
39 | PdfContentReaderFlags::SkipHandleNonFormXObjects;
40
41 PdfContentStreamReader reader(device, args);
42 PdfContent content;
43
44 // Must terminate — before the fix this loops forever.
45 // A generous iteration cap ensures the test fails fast if
46 // the bug regresses, without relying on wall-clock timeouts.
47 unsigned iterations = 0;
48 constexpr unsigned maxIterations = 100;
49 try
50 {
51 while (reader.TryReadNext(content) && iterations < maxIterations)
52 iterations++;
53 }
54 catch (PdfError&)
55 {
56 }
57 catch (std::exception&)
58 {
59 }
60
61 REQUIRE(iterations < maxIterations);
62 }
63 }