| 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 |
// Subclass PdfTokenizer to expose protected ReadString / ReadHexString |
| 10 |
class TestableTokenizer : public PdfTokenizer |
| 11 |
{ |
| 12 |
public: |
| 13 |
using PdfTokenizer::ReadString; |
| 14 |
using PdfTokenizer::ReadHexString; |
| 15 |
}; |
| 16 |
|
| 17 |
// --------------------------------------------------------------------------- |
| 18 |
// 1) Unterminated regular string – must be rejected when it exceeds the |
| 19 |
// maximum allowed string length instead of growing until OOM. |
| 20 |
// --------------------------------------------------------------------------- |
| 21 |
TEST_CASE("TestStringGrowthIsBounded") |
| 22 |
{ |
| 23 |
// Build a '(' followed by data exceeding MaxStringLength with no |
| 24 |
// closing ')'. After the fix the size cap rejects this; before the |
| 25 |
// fix m_charBuffer grows without limit. |
| 26 |
constexpr size_t payloadSize = PdfTokenizer::MaxStringLength + 1; |
| 27 |
string input(payloadSize + 1, 'A'); |
| 28 |
input[0] = '('; // opening paren, never closed |
| 29 |
|
| 30 |
SpanStreamDevice device(input); |
| 31 |
TestableTokenizer tokenizer; |
| 32 |
PdfVariant variant; |
| 33 |
|
| 34 |
// Consume the '(' token first |
| 35 |
string_view token; |
| 36 |
PdfTokenType tokenType; |
| 37 |
REQUIRE(tokenizer.TryReadNextToken(device, token, tokenType)); |
| 38 |
REQUIRE(tokenType == PdfTokenType::ParenthesisLeft); |
| 39 |
|
| 40 |
// ReadString should throw ValueOutOfRange because the string exceeds |
| 41 |
// the maximum allowed length |
| 42 |
try |
| 43 |
{ |
| 44 |
tokenizer.ReadString(device, variant, nullptr, { true, false }); |
| 45 |
FAIL("Expected PdfError to be thrown"); |
| 46 |
} |
| 47 |
catch (const PdfError& e) |
| 48 |
{ |
| 49 |
REQUIRE(e.GetCode() == PdfErrorCode::ValueOutOfRange); |
| 50 |
} |
| 51 |
} |
| 52 |
|
| 53 |
// --------------------------------------------------------------------------- |
| 54 |
// 2) Unterminated hex string – same pattern, different code path. |
| 55 |
// --------------------------------------------------------------------------- |
| 56 |
TEST_CASE("TestHexStringGrowthIsBounded") |
| 57 |
{ |
| 58 |
constexpr size_t payloadSize = PdfTokenizer::MaxStringLength + 1; |
| 59 |
string input(payloadSize + 1, 'A'); // 'A' is a valid hex digit |
| 60 |
input[0] = '<'; // opening angle bracket, never closed |
| 61 |
|
| 62 |
SpanStreamDevice device(input); |
| 63 |
TestableTokenizer tokenizer; |
| 64 |
PdfVariant variant; |
| 65 |
|
| 66 |
string_view token; |
| 67 |
PdfTokenType tokenType; |
| 68 |
REQUIRE(tokenizer.TryReadNextToken(device, token, tokenType)); |
| 69 |
REQUIRE(tokenType == PdfTokenType::AngleBracketLeft); |
| 70 |
|
| 71 |
try |
| 72 |
{ |
| 73 |
tokenizer.ReadHexString(device, variant, nullptr, { true, false }); |
| 74 |
FAIL("Expected PdfError to be thrown"); |
| 75 |
} |
| 76 |
catch (const PdfError& e) |
| 77 |
{ |
| 78 |
REQUIRE(e.GetCode() == PdfErrorCode::ValueOutOfRange); |
| 79 |
} |
| 80 |
} |
| 81 |
|
| 82 |
// --------------------------------------------------------------------------- |
| 83 |
// 3) Object stream with offset beyond buffer – readObjectsFromStream must |
| 84 |
// reject first+offset that exceeds the decompressed buffer length. |
| 85 |
// We test via PdfObjectStreamParser indirectly by crafting a minimal PDF |
| 86 |
// with an ObjStm whose /First is larger than the stream content. |
| 87 |
// |
| 88 |
// This test builds a complete XRef-stream-based PDF so the parser |
| 89 |
// actually enters the object stream parsing code path. |
| 90 |
// --------------------------------------------------------------------------- |
| 91 |
TEST_CASE("TestObjectStreamOffsetBeyondBuffer") |
| 92 |
{ |
| 93 |
// Strategy: Build a PDF 1.5 document with: |
| 94 |
// obj 1 = catalog (normal object at known offset) |
| 95 |
// obj 2 = ObjStm with /First 99999, but stream content is tiny |
| 96 |
// obj 3 = compressed inside obj 2 (referenced via XRef stream type-2) |
| 97 |
// obj 4 = XRef stream (the cross-reference itself) |
| 98 |
// |
| 99 |
// The XRef stream tells the parser that obj 3 lives inside obj 2, |
| 100 |
// triggering PdfObjectStreamParser::Parse which should now fail |
| 101 |
// because /First (99999) exceeds the stream buffer. |
| 102 |
|
| 103 |
// --- Object 1: catalog --- |
| 104 |
ostringstream oss; |
| 105 |
oss << "%PDF-1.5\n"; |
| 106 |
|
| 107 |
size_t obj1Pos = oss.str().size(); |
| 108 |
oss << "1 0 obj\n" |
| 109 |
<< "<< /Type /Catalog /Pages 1 0 R >>\n" |
| 110 |
<< "endobj\n"; |
| 111 |
|
| 112 |
// --- Object 2: ObjStm with bad /First --- |
| 113 |
// Stream content: "3 0\n42" means object #3 is at offset 0 from /First, value=42 |
| 114 |
string objStmContent = "3 0\n42"; |
| 115 |
size_t obj2Pos = oss.str().size(); |
| 116 |
oss << "2 0 obj\n" |
| 117 |
<< "<< /Type /ObjStm /N 1 /First 99999 /Length " |
| 118 |
<< objStmContent.size() << " >>\n" |
| 119 |
<< "stream\n" |
| 120 |
<< objStmContent |
| 121 |
<< "\nendstream\n" |
| 122 |
<< "endobj\n"; |
| 123 |
|
| 124 |
// --- Object 4: XRef stream --- |
| 125 |
// W = [1 2 1]: type(1 byte), field2(2 bytes), field3(1 byte) |
| 126 |
// Entries (5 total: obj 0-4): |
| 127 |
// obj 0: type=0 (free), next=0, gen=0 |
| 128 |
// obj 1: type=1 (normal), offset=obj1Pos, gen=0 |
| 129 |
// obj 2: type=1 (normal), offset=obj2Pos, gen=0 |
| 130 |
// obj 3: type=2 (compressed), stream_obj=2, index=0 |
| 131 |
// obj 4: type=1 (normal), offset=obj4Pos, gen=0 <-- self |
| 132 |
size_t obj4Pos = oss.str().size(); |
| 133 |
|
| 134 |
// We need to encode the xref stream binary content. |
| 135 |
// Each entry is 4 bytes: [type:1][field2_hi:1][field2_lo:1][field3:1] |
| 136 |
// We'll use ASCIIHexDecode so we can write hex in the stream. |
| 137 |
// Entry format with W=[1 2 1]: |
| 138 |
// obj 0: 00 0000 00 |
| 139 |
// obj 1: 01 XXXX 00 (XXXX = obj1Pos as 2 bytes) |
| 140 |
// obj 2: 01 YYYY 00 (YYYY = obj2Pos as 2 bytes) |
| 141 |
// obj 3: 02 0002 00 (compressed in obj 2, index 0) |
| 142 |
// obj 4: 01 ZZZZ 00 (ZZZZ = obj4Pos as 2 bytes) |
| 143 |
|
| 144 |
ostringstream xrefData; |
| 145 |
xrefData << utls::Format("{:02X}{:04X}{:02X}", 0, 0, 0); |
| 146 |
xrefData << utls::Format("{:02X}{:04X}{:02X}", 1, (unsigned)obj1Pos, 0); |
| 147 |
xrefData << utls::Format("{:02X}{:04X}{:02X}", 1, (unsigned)obj2Pos, 0); |
| 148 |
xrefData << utls::Format("{:02X}{:04X}{:02X}", 2, 2, 0); |
| 149 |
xrefData << utls::Format("{:02X}{:04X}{:02X}", 1, (unsigned)obj4Pos, 0); |
| 150 |
string xrefHex = xrefData.str(); |
| 151 |
|
| 152 |
oss << "4 0 obj\n" |
| 153 |
<< "<< /Type /XRef /Size 5 /W [1 2 1]" |
| 154 |
<< " /Root 1 0 R" |
| 155 |
<< " /Filter /ASCIIHexDecode" |
| 156 |
<< " /Length " << xrefHex.size() |
| 157 |
<< " >>\n" |
| 158 |
<< "stream\n" |
| 159 |
<< xrefHex |
| 160 |
<< "\nendstream\n" |
| 161 |
<< "endobj\n"; |
| 162 |
|
| 163 |
oss << "startxref\n" << obj4Pos << "\n%%EOF\n"; |
| 164 |
|
| 165 |
string pdfBuf = oss.str(); |
| 166 |
|
| 167 |
// Loading should trigger reading obj 3 from the ObjStm (obj 2), |
| 168 |
// which should fail because /First=99999 exceeds the stream content. |
| 169 |
// Before the fix, this is caught incidentally by SpanStreamDevice::Seek |
| 170 |
// throwing ValueOutOfRange. After the fix, PdfObjectStreamParser should |
| 171 |
// detect this explicitly and throw BrokenFile before reaching the Seek. |
| 172 |
bool threw = false; |
| 173 |
PdfErrorCode caughtCode = PdfErrorCode::Unknown; |
| 174 |
try |
| 175 |
{ |
| 176 |
PdfMemDocument doc; |
| 177 |
doc.LoadFromBuffer(pdfBuf); |
| 178 |
// If we get here, access obj 3 to force parsing |
| 179 |
(void)doc.GetObjects().GetObject(PdfReference(3, 0)); |
| 180 |
} |
| 181 |
catch (const PdfError& e) |
| 182 |
{ |
| 183 |
threw = true; |
| 184 |
caughtCode = e.GetCode(); |
| 185 |
} |
| 186 |
REQUIRE(threw); |
| 187 |
// After the fix this should be BrokenFile (explicit bounds check); |
| 188 |
// before the fix it was ValueOutOfRange (incidental Seek failure). |
| 189 |
REQUIRE(caughtCode == PdfErrorCode::BrokenFile); |
| 190 |
} |
| 191 |
|
| 192 |
|