deleted file mode 100644
@@ -1,225 +0,0 @@
-From e974684acbe3aeb6bc6192b06a048537c345b713 Mon Sep 17 00:00:00 2001
-From: Stefan Weil <sw@weilnetz.de>
-Date: Mon, 24 Aug 2026 13:39:22 +0200
-Subject: [PATCH] Limit unichar extraction in ReadNormProtos to the buffer size
-
-Classify::ReadNormProtos parsed each normproto line with
-`stream >> unichar >> NumProtos` into a char unichar[2 * UNICHAR_LEN + 1]
-stack buffer, but char* extraction from an istream has no length limit
-(the stream width was never set). A crafted TESSDATA_NORMPROTO component
-in a .traineddata file whose first proto-line token exceeds 60
-characters (the 100-byte line buffer allows up to 99) overflows the
-stack buffer during legacy engine initialization (CWE-121).
-
-Toolchain note: Apple's libc++ provides a C++20 array overload of
-operator>>(basic_istream&, char(&)[N]) that implicitly bounds the
-extraction to the array size, so builds against that standard library
-are incidentally protected. Standard libraries without that overload
-(e.g. libstdc++) still take the unbounded char* overload, so the
-explicit width limit below makes the behavior defined on all
-toolchains.
-
-Key changes:
-- normmatch.cpp: read the unichar token with
- std::setw(2 * UNICHAR_LEN + 1); char* extraction takes at most
- width - 1 characters, which fits the buffer exactly. Overlong
- tokens are truncated and the line is rejected like any other
- unparseable line.
-- unittest: add normproto_test covering a 99-character token (the
- maximum a 100-byte line can hold), a token of exactly 2 *
- UNICHAR_LEN characters, and a well-formed component. A minimal
- reproduction of the unbounded extraction crashes an ASan build
- with a stack-buffer-overflow.
-
-Reported-by: Tristan Madani <tristan@talencesecurity.com>
-Assisted-by: OpenCode / qwen3.8-27b-thinking (Alibaba Cloud)
-Signed-off-by: Stefan Weil <sw@weilnetz.de>
-(cherry picked from commit 1bda5079b1c8a7e25f523486837426903d29ce84)
-
-CVE: CVE-2026-88047
-Upstream-Status: Backport [https://github.com/tesseract-ocr/tesseract/commit/1bda5079b1c8a7e25f523486837426903d29ce84]
-Signed-off-by: Ankur Tyagi <ankur.tyagi85@gmail.com>
----
- Makefile.am | 5 ++
- src/classify/normmatch.cpp | 6 +-
- unittest/CMakeLists.txt | 1 +
- unittest/normproto_test.cc | 111 +++++++++++++++++++++++++++++++++++++
- 4 files changed, 122 insertions(+), 1 deletion(-)
- create mode 100644 unittest/normproto_test.cc
-
-diff --git a/Makefile.am b/Makefile.am
-index 63ca6686..cc748455 100644
---- a/Makefile.am
-+++ b/Makefile.am
-@@ -1184,6 +1184,7 @@ check_PROGRAMS += mastertrainer_test
- endif # !DISABLED_LEGACY_ENGINE
- check_PROGRAMS += matrix_test
- check_PROGRAMS += networkio_test
-+check_PROGRAMS += normproto_test
- if ENABLE_TRAINING
- check_PROGRAMS += normstrngs_test
- endif # ENABLE_TRAINING
-@@ -1388,6 +1389,10 @@ networkio_test_SOURCES = unittest/networkio_test.cc
- networkio_test_CPPFLAGS = $(unittest_CPPFLAGS)
- networkio_test_LDADD = $(TESS_LIBS)
-
-+normproto_test_SOURCES = unittest/normproto_test.cc
-+normproto_test_CPPFLAGS = $(unittest_CPPFLAGS)
-+normproto_test_LDADD = $(TESS_LIBS)
-+
- normstrngs_test_SOURCES = unittest/normstrngs_test.cc
- normstrngs_test_CPPFLAGS = $(unittest_CPPFLAGS)
- normstrngs_test_LDADD = $(TRAINING_LIBS) $(ICU_I18N_LIBS) $(ICU_UC_LIBS)
-diff --git a/src/classify/normmatch.cpp b/src/classify/normmatch.cpp
-index d5bd7e6a..1ce7c928 100644
---- a/src/classify/normmatch.cpp
-+++ b/src/classify/normmatch.cpp
-@@ -28,6 +28,7 @@
-
- #include <cmath>
- #include <cstdio>
-+#include <iomanip> // for std::setw
- #include <sstream> // for std::istringstream
-
- namespace tesseract {
-@@ -190,7 +191,10 @@ NORM_PROTOS *Classify::ReadNormProtos(TFile *fp) {
- while (fp->FGets(line, kMaxLineSize) != nullptr) {
- std::istringstream stream(line);
- stream.imbue(std::locale::classic());
-- stream >> unichar >> NumProtos;
-+ // unichar holds at most 2 * UNICHAR_LEN characters; the width limit
-+ // (width - 1 characters for char* extraction) keeps the extraction
-+ // from overflowing the buffer on overlong lines.
-+ stream >> std::setw(2 * UNICHAR_LEN + 1) >> unichar >> NumProtos;
- if (stream.fail()) {
- continue;
- }
-diff --git a/unittest/CMakeLists.txt b/unittest/CMakeLists.txt
-index a660c9b1..4c677a60 100644
---- a/unittest/CMakeLists.txt
-+++ b/unittest/CMakeLists.txt
-@@ -59,6 +59,7 @@ set(LEGACY_TESTS
- indexmapbidi_test.cc
- intfeaturemap_test.cc
- mastertrainer_test.cc
-+ normproto_test.cc
- osd_test.cc
- params_model_test.cc
- shapetable_test.cc
-diff --git a/unittest/normproto_test.cc b/unittest/normproto_test.cc
-new file mode 100644
-index 00000000..574b2f3d
---- /dev/null
-+++ b/unittest/normproto_test.cc
-@@ -0,0 +1,111 @@
-+///////////////////////////////////////////////////////////////////////
-+// File: normproto_test.cc
-+// Description: Tests that Classify::ReadNormProtos handles a normproto
-+// line whose first (unichar) token exceeds the
-+// unichar[2 * UNICHAR_LEN + 1] stack buffer. The
-+// istream extraction has no intrinsic length limit, so a
-+// crafted NORMPROTO component in a .traineddata file
-+// could overflow the stack buffer during legacy engine
-+// initialization.
-+//
-+// Licensed under the Apache License, Version 2.0 (the "License");
-+// you may not use this file except in compliance with the License.
-+// You may obtain a copy of the License at
-+// http://www.apache.org/licenses/LICENSE-2.0
-+//
-+///////////////////////////////////////////////////////////////////////
-+
-+#include "include_gunit.h"
-+
-+#include "classify.h"
-+#include "serialis.h" // for TFile
-+
-+#include <cstdio>
-+#include <cstdlib>
-+#include <string>
-+#include <vector>
-+
-+namespace tesseract {
-+namespace {
-+
-+// Minimal unicharset (space and 'a').
-+const char kMinUnicharset[] =
-+ "2\n"
-+ "NULL 1 0,255,0,255,0,0,0,0,0,0 Latin 2 0 2\n"
-+ "a 1 0,255,0,255,0,0,0,0,0,0 Latin 2 0 2\n";
-+
-+// Builds a normproto component: a sample-size line (5), five parameter
-+// description lines, then the given raw proto lines.
-+std::vector<char> MakeNormproto(const std::string &lines) {
-+ std::string data = "5\n";
-+ for (int i = 0; i < 5; ++i) {
-+ data += "e e 0 1\n";
-+ }
-+ data += lines;
-+ return std::vector<char>(data.begin(), data.end());
-+}
-+
-+class NormprotoTest : public testing::Test {
-+protected:
-+ void SetUp() override {
-+ tmpl_ = "/tmp/tess_normproto_test_XXXXXX";
-+ char *dir = mkdtemp(tmpl_.data());
-+ ASSERT_NE(dir, nullptr);
-+ dir_ = dir;
-+ std::string uc_path = dir_ + "/eng.unicharset";
-+ FILE *f = fopen(uc_path.c_str(), "w");
-+ ASSERT_NE(f, nullptr);
-+ ASSERT_EQ(fwrite(kMinUnicharset, 1, sizeof(kMinUnicharset) - 1, f),
-+ sizeof(kMinUnicharset) - 1);
-+ fclose(f);
-+ // Load the minimal unicharset into the classifier's inherited
-+ // unicharset member.
-+ ASSERT_TRUE(classifier_.unicharset.load_from_file(uc_path.c_str()));
-+ }
-+ void TearDown() override {
-+ std::remove((dir_ + "/eng.unicharset").c_str());
-+ rmdir(dir_.c_str());
-+ }
-+ std::string dir_;
-+ std::string tmpl_;
-+ Classify classifier_;
-+};
-+
-+// A 99-character first token (the maximum FGets can return) overflows
-+// unichar[2 * UNICHAR_LEN + 1] on unpatched code; the width-limited
-+// extraction must reject the line instead.
-+TEST_F(NormprotoTest, ToleratesOverlongUnicharToken) {
-+ std::vector<char> bytes = MakeNormproto(std::string(99, 'A') + "\n");
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ classifier_.NormProtos = classifier_.ReadNormProtos(&fp);
-+ ASSERT_NE(classifier_.NormProtos, nullptr);
-+ classifier_.FreeNormProtos();
-+ EXPECT_EQ(classifier_.NormProtos, nullptr);
-+}
-+
-+// A token of exactly 2 * UNICHAR_LEN characters is the maximum legitimate
-+// size; it must not be truncated or rejected by the width limit.
-+TEST_F(NormprotoTest, ToleratesMaxLenUnicharToken) {
-+ std::vector<char> bytes = MakeNormproto(std::string(2 * UNICHAR_LEN, 'A') + " 0\n");
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ classifier_.NormProtos = classifier_.ReadNormProtos(&fp);
-+ ASSERT_NE(classifier_.NormProtos, nullptr);
-+ classifier_.FreeNormProtos();
-+ EXPECT_EQ(classifier_.NormProtos, nullptr);
-+}
-+
-+// A well-formed normproto component must still parse.
-+TEST_F(NormprotoTest, ReadsValidNormprotos) {
-+ std::vector<char> bytes = MakeNormproto("a 0\n");
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ classifier_.NormProtos = classifier_.ReadNormProtos(&fp);
-+ ASSERT_NE(classifier_.NormProtos, nullptr);
-+ classifier_.FreeNormProtos();
-+ EXPECT_EQ(classifier_.NormProtos, nullptr);
-+}
-+
-+} // namespace
-+} // namespace tesseract
deleted file mode 100644
@@ -1,227 +0,0 @@
-From b2fbc8282c34f0a5a426622a9eb6a79ad0df0fe4 Mon Sep 17 00:00:00 2001
-From: Stefan Weil <sw@weilnetz.de>
-Date: Fri, 21 Aug 2026 15:23:59 +0200
-Subject: [PATCH] Validate FullyConnected weight matrix dimensions at load
-
-FullyConnected::DeSerialize read the WeightMatrix without checking
-that its dimensions match the layer's declared ni/no. MatrixDotVector
-drives the dot product from the matrix dimensions (writes w.dim1()
-results, reads w.dim2()-1 inputs) while the scratch buffers are sized
-from no_ and ni_, so a crafted .traineddata with a mismatched matrix
-performed a heap out-of-bounds write (up to 65535 rows) and out-of-
-bounds read on the first recognition step (crash, or heap corruption
-with attacker-influenced size and content).
-
-Key changes:
-- weightmatrix.h: add Dim1()/Dim2() accessors for the active weight
- matrix (int or float), alongside the existing NumOutputs().
-- fullyconnected.cpp: reject the layer in DeSerialize unless
- Dim1() == no_ and Dim2() == ni_ + 1 (the second dimension
- includes the bias column), the layout that InitWeightsFloat
- always produces.
-- unittest: new fullyconnected_test that builds a Softmax network
- with mismatched matrix dimensions and expects CreateFromFile to
- return nullptr; on unpatched code the test reaches Forward and
- ASan catches the out-of-bounds write in MatrixDotVector. A
- second test verifies that matching dimensions are still accepted.
-
-Reported-by: Zhixi "Jace" Sun <g.mygenie@gmail.com>
-Assisted-by: OpenCode / qwen3.8-27b-thinking (Alibaba Cloud)
-Signed-off-by: Stefan Weil <sw@weilnetz.de>
-(cherry picked from commit 103dc134eb36411ddc6833ec20aa2c76795bd0ff)
-
-CVE: CVE-2026-88048
-Upstream-Status: Backport [https://github.com/tesseract-ocr/tesseract/commit/103dc134eb36411ddc6833ec20aa2c76795bd0ff]
-Signed-off-by: Ankur Tyagi <ankur.tyagi85@gmail.com>
----
- Makefile.am | 5 ++
- src/lstm/fullyconnected.cpp | 11 ++-
- src/lstm/weightmatrix.h | 7 ++
- unittest/fullyconnected_test.cc | 115 ++++++++++++++++++++++++++++++++
- 4 files changed, 137 insertions(+), 1 deletion(-)
- create mode 100644 unittest/fullyconnected_test.cc
-
-diff --git a/Makefile.am b/Makefile.am
-index 22c2c7a0..284f8b53 100644
---- a/Makefile.am
-+++ b/Makefile.am
-@@ -1158,6 +1158,7 @@ if !DISABLED_LEGACY_ENGINE
- check_PROGRAMS += equationdetect_test
- endif # !DISABLED_LEGACY_ENGINE
- check_PROGRAMS += fileio_test
-+check_PROGRAMS += fullyconnected_test
- check_PROGRAMS += heap_test
- check_PROGRAMS += imagedata_test
- if !DISABLED_LEGACY_ENGINE
-@@ -1290,6 +1291,10 @@ fileio_test_SOURCES = unittest/fileio_test.cc
- fileio_test_CPPFLAGS = $(unittest_CPPFLAGS)
- fileio_test_LDADD = $(TRAINING_LIBS)
-
-+fullyconnected_test_SOURCES = unittest/fullyconnected_test.cc
-+fullyconnected_test_CPPFLAGS = $(unittest_CPPFLAGS)
-+fullyconnected_test_LDADD = $(TESS_LIBS)
-+
- heap_test_SOURCES = unittest/heap_test.cc
- heap_test_CPPFLAGS = $(unittest_CPPFLAGS)
- heap_test_LDADD = $(TESS_LIBS)
-diff --git a/src/lstm/fullyconnected.cpp b/src/lstm/fullyconnected.cpp
-index a0ee1b02..d999f575 100644
---- a/src/lstm/fullyconnected.cpp
-+++ b/src/lstm/fullyconnected.cpp
-@@ -121,7 +121,16 @@ bool FullyConnected::Serialize(TFile *fp) const {
-
- // Reads from the given file. Returns false in case of error.
- bool FullyConnected::DeSerialize(TFile *fp) {
-- return weights_.DeSerialize(IsTraining(), fp);
-+ if (!weights_.DeSerialize(IsTraining(), fp)) {
-+ return false;
-+ }
-+ // The weight matrix must match the declared sizes (the second dimension
-+ // includes the bias column); otherwise Forward would read or write
-+ // outside the scratch buffers sized from ni_ and no_.
-+ if (weights_.Dim1() != no_ || weights_.Dim2() != ni_ + 1) {
-+ return false;
-+ }
-+ return true;
- }
-
- // Runs forward propagation of activations on the input line.
-diff --git a/src/lstm/weightmatrix.h b/src/lstm/weightmatrix.h
-index eaca3ffb..6c5ab236 100644
---- a/src/lstm/weightmatrix.h
-+++ b/src/lstm/weightmatrix.h
-@@ -107,6 +107,13 @@ public:
- int NumOutputs() const {
- return int_mode_ ? wi_.dim1() : wf_.dim1();
- }
-+ // The dimensions of the active weight matrix (wi_ in int mode, else wf_).
-+ int Dim1() const {
-+ return int_mode_ ? wi_.dim1() : wf_.dim1();
-+ }
-+ int Dim2() const {
-+ return int_mode_ ? wi_.dim2() : wf_.dim2();
-+ }
- // Provides one set of weights. Only used by peep weight maxpool.
- const TFloat *GetWeights(int index) const {
- return wf_[index];
-diff --git a/unittest/fullyconnected_test.cc b/unittest/fullyconnected_test.cc
-new file mode 100644
-index 00000000..f9697de4
---- /dev/null
-+++ b/unittest/fullyconnected_test.cc
-@@ -0,0 +1,115 @@
-+///////////////////////////////////////////////////////////////////////
-+// File: fullyconnected_test.cc
-+// Description: Tests that a FullyConnected (softmax) network layer with
-+// weight-matrix dimensions that do not match the declared
-+// ni/no is rejected at load. Without the check,
-+// MatrixDotVector writes w.dim1() results into a scratch
-+// buffer sized from no_ and reads w.dim2()-1 inputs from
-+// a buffer sized from ni_ (heap out-of-bounds write/read)
-+// on the first recognition step.
-+//
-+// Licensed under the Apache License, Version 2.0 (the "License");
-+// you may not use this file except in compliance with the License.
-+// You may obtain a copy of the License at
-+// http://www.apache.org/licenses/LICENSE-2.0
-+//
-+///////////////////////////////////////////////////////////////////////
-+
-+#include "include_gunit.h"
-+
-+#include "network.h" // for Network, NetworkType
-+#include "networkio.h"
-+#include "networkscratch.h"
-+#include "serialis.h" // for TFile
-+
-+#include <cstdint>
-+#include <vector>
-+
-+namespace tesseract {
-+namespace {
-+
-+// Appends raw little-endian values to a byte buffer.
-+class ByteWriter {
-+public:
-+ void PutU8(uint32_t v) { data_.push_back(static_cast<char>(v & 0xFF)); }
-+ void PutU32(uint32_t v) {
-+ for (int i = 0; i < 4; ++i) {
-+ data_.push_back(static_cast<char>((v >> (8 * i)) & 0xFF));
-+ }
-+ }
-+ void PutS32(int32_t v) { PutU32(static_cast<uint32_t>(v)); }
-+ const std::vector<char> &data() const { return data_; }
-+
-+private:
-+ std::vector<char> data_;
-+};
-+
-+void PutDoubleLE(ByteWriter *w, double d) {
-+ union {
-+ double d;
-+ uint64_t u;
-+ } conv;
-+ conv.d = d;
-+ w->PutU32(static_cast<uint32_t>(conv.u & 0xFFFFFFFF));
-+ w->PutU32(static_cast<uint32_t>(conv.u >> 32));
-+}
-+
-+// Builds a serialized NT_SOFTMAX network: header with the given ni/no,
-+// then a float-mode WeightMatrix with the given (corrupt) dimensions.
-+// The matrix is stored as doubles on disk (see WeightMatrix::DeSerialize).
-+std::vector<char> MakeSoftmaxNetwork(int ni, int no, int32_t dim1, int32_t dim2) {
-+ ByteWriter w;
-+ w.PutU8(static_cast<uint32_t>(NT_SOFTMAX));
-+ w.PutU8(0); // training: TS_DISABLED
-+ w.PutU8(0); // needs_to_backprop
-+ w.PutU32(0); // network_flags
-+ w.PutU32(static_cast<uint32_t>(ni));
-+ w.PutU32(static_cast<uint32_t>(no));
-+ w.PutU32(0); // num_weights (not cross-checked, kept consistent anyway)
-+ w.PutU32(0); // name (empty string)
-+ // WeightMatrix::DeSerialize:
-+ w.PutU8(128); // mode: kDoubleFlag, float mode
-+ w.PutS32(dim1);
-+ w.PutS32(dim2);
-+ PutDoubleLE(&w, 0.0); // empty_ cell
-+ for (int32_t i = 0; i < dim1 * dim2; ++i) {
-+ PutDoubleLE(&w, 0.0); // weight data
-+ }
-+ return w.data();
-+}
-+
-+// A FullyConnected layer whose weight matrix does not match the declared
-+// sizes must be rejected by CreateFromFile; on unpatched code the test
-+// reaches Forward, where MatrixDotVector performs the out-of-bounds
-+// write this regression test guards against.
-+TEST(FullyconnectedTest, RejectsWeightMatrixDimensionMismatch) {
-+ // ni_=no_=1 but dim1=3 (OOB write of 3 results into a 1-result buffer)
-+ // and dim2=5 (OOB read of 4 inputs from a 1-input buffer).
-+ std::vector<char> bytes = MakeSoftmaxNetwork(1, 1, 3, 5);
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ Network *net = Network::CreateFromFile(&fp);
-+ if (net == nullptr) {
-+ return; // Fixed: the mismatched layer is rejected at load.
-+ }
-+ NetworkIO input;
-+ input.Resize2d(false, /*width=*/1, /*num_features=*/1);
-+ NetworkScratch scratch;
-+ NetworkIO output;
-+ net->Forward(false, input, nullptr, &scratch, &output);
-+ delete net;
-+ FAIL() << "crafted FullyConnected layer with mismatched weight matrix was accepted";
-+}
-+
-+// Consistent dimensions must still be accepted.
-+TEST(FullyconnectedTest, AcceptsMatchingDimensions) {
-+ std::vector<char> bytes = MakeSoftmaxNetwork(1, 2, 2, 2);
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ Network *net = Network::CreateFromFile(&fp);
-+ ASSERT_NE(net, nullptr);
-+ delete net;
-+}
-+
-+} // namespace
-+} // namespace tesseract
deleted file mode 100644
@@ -1,315 +0,0 @@
-From 62fabbca097d127e7d5a8fe3869102d2e4918b80 Mon Sep 17 00:00:00 2001
-From: Stefan Weil <sw@weilnetz.de>
-Date: Fri, 21 Aug 2026 16:41:51 +0200
-Subject: [PATCH] Validate LSTM gate matrix dimensions against na_/no_ at load
-
-LSTM::DeSerialize read na_ from the untrusted TESSDATA_LSTM component
-and derived ns_ from the CI gate matrix's dim1, but never checked that
-the deserialized dimensions were mutually consistent. The forward pass
-sizes its buffers from na_, no_ and ns_ while the gate matrices drive
-their own dimensions, so a crafted .traineddata performed heap
-out-of-bounds writes and reads during the first recognition step, e.g.
-WriteTimeStepPart writing ns_ floats at offset ni_+nf_ into a source_
-buffer sized from na_ (up to ~256 KB with an attacker-chosen gate
-matrix dim1).
-
-The bounds assertions added by 2f4d2f4 (CVE-2026-73066) covered only
-NetworkIO::CopyTimeStepGeneral and Randomize, leaving
-WriteTimeStepPart and AddTimeStepPart unguarded.
-
-Key changes:
-- weightmatrix.h: add Dim1()/Dim2() accessors for the active weight
- matrix (int or float), alongside the existing NumOutputs().
-- lstm.cpp: reject the layer in DeSerialize unless na_ ==
- ni_ + nf_ + (is_2d_ ? 2 : 1) * ns_, every deserialized gate has
- Dim1() == ns_ and Dim2() == na_ + 1 (the layout InitWeightsFloat
- always produces), ns_ == no_ for plain NT_LSTM/NT_LSTM_SUMMARY, and
- the softmax layer's sizes match ns_/no_ for the softmax variants.
-- networkio.cpp: add the same defense-in-depth bounds assertions to
- WriteTimeStepPart and AddTimeStepPart as 2f4d2f4 added to
- CopyTimeStepGeneral and Randomize.
-- unittest: add lstm_layer_test with crafted NT_LSTM layers for the
- na_ mismatch, gate dim1 mismatch, and gate dim2 mismatch cases
- (each rejected at load; on unpatched code the tests reach Forward
- and ASan catches the out-of-bounds write in WriteTimeStepPart),
- plus a positive control that a consistent layer loads and runs.
-
-Reported-by: Zhixi "Jace" Sun <g.mygenie@gmail.com>
-Assisted-by: OpenCode / qwen3.8-27b-thinking (Alibaba Cloud)
-Signed-off-by: Stefan Weil <sw@weilnetz.de>
-(cherry picked from commit b494ac18925f9d9aff9ef5815475de9943ab19bf)
-
-CVE: CVE-2026-88049
-Upstream-Status: Backport [https://github.com/tesseract-ocr/tesseract/commit/b494ac18925f9d9aff9ef5815475de9943ab19bf]
-Signed-off-by: Ankur Tyagi <ankur.tyagi85@gmail.com>
----
- Makefile.am | 5 +
- src/lstm/lstm.cpp | 20 ++++
- src/lstm/networkio.cpp | 2 +
- unittest/lstm_layer_test.cc | 177 ++++++++++++++++++++++++++++++++++++
- 4 files changed, 204 insertions(+)
- create mode 100644 unittest/lstm_layer_test.cc
-
-diff --git a/Makefile.am b/Makefile.am
-index 284f8b53..4bcc858b 100644
---- a/Makefile.am
-+++ b/Makefile.am
-@@ -1171,6 +1171,7 @@ check_PROGRAMS += layout_test
- check_PROGRAMS += ligature_table_test
- check_PROGRAMS += linlsq_test
- check_PROGRAMS += list_test
-+check_PROGRAMS += lstm_layer_test
- if ENABLE_TRAINING
- check_PROGRAMS += lstm_recode_test
- check_PROGRAMS += lstm_squashed_test
-@@ -1352,6 +1353,10 @@ loadlang_test_SOURCES = unittest/loadlang_test.cc
- loadlang_test_CPPFLAGS = $(unittest_CPPFLAGS)
- loadlang_test_LDADD = $(TESS_LIBS) $(LEPTONICA_LIBS)
-
-+lstm_layer_test_SOURCES = unittest/lstm_layer_test.cc
-+lstm_layer_test_CPPFLAGS = $(unittest_CPPFLAGS)
-+lstm_layer_test_LDADD = $(TESS_LIBS)
-+
- lstm_recode_test_SOURCES = unittest/lstm_recode_test.cc
- lstm_recode_test_CPPFLAGS = $(unittest_CPPFLAGS)
- lstm_recode_test_LDADD = $(TRAINING_LIBS)
-diff --git a/src/lstm/lstm.cpp b/src/lstm/lstm.cpp
-index 7bfd19b5..bfffa5de 100644
---- a/src/lstm/lstm.cpp
-+++ b/src/lstm/lstm.cpp
-@@ -274,12 +274,32 @@ bool LSTM::DeSerialize(TFile *fp) {
- is_2d_ = na_ - nf_ == ni_ + 2 * ns_;
- }
- }
-+ // The deserialized dimensions must be mutually consistent: the forward
-+ // pass sizes its buffers from na_, no_ and ns_ while the gate matrices
-+ // drive their own dimensions.
-+ if (na_ != ni_ + nf_ + (is_2d_ ? 2 : 1) * ns_) {
-+ return false;
-+ }
-+ for (int w = 0; w < WT_COUNT; ++w) {
-+ if (w == GFS && !Is2D()) {
-+ continue;
-+ }
-+ if (gate_weights_[w].Dim1() != ns_ || gate_weights_[w].Dim2() != na_ + 1) {
-+ return false;
-+ }
-+ }
-+ if ((type_ == NT_LSTM || type_ == NT_LSTM_SUMMARY) && ns_ != no_) {
-+ return false;
-+ }
- delete softmax_;
- if (type_ == NT_LSTM_SOFTMAX || type_ == NT_LSTM_SOFTMAX_ENCODED) {
- softmax_ = static_cast<FullyConnected *>(Network::CreateFromFile(fp));
- if (softmax_ == nullptr) {
- return false;
- }
-+ if (softmax_->NumInputs() != ns_ || softmax_->NumOutputs() != no_) {
-+ return false;
-+ }
- } else {
- softmax_ = nullptr;
- }
-diff --git a/src/lstm/networkio.cpp b/src/lstm/networkio.cpp
-index 8e1c6679..7a31034b 100644
---- a/src/lstm/networkio.cpp
-+++ b/src/lstm/networkio.cpp
-@@ -641,6 +641,7 @@ void NetworkIO::AddTimeStep(int t, TFloat *inout) const {
-
- // Adds part of a single timestep to floats.
- void NetworkIO::AddTimeStepPart(int t, int offset, int num_features, float *inout) const {
-+ ASSERT_HOST(offset + num_features <= NumFeatures());
- if (int_mode_) {
- const int8_t *line = i_[t] + offset;
- for (int i = 0; i < num_features; ++i) {
-@@ -662,6 +663,7 @@ void NetworkIO::WriteTimeStep(int t, const TFloat *input) {
- // Writes a single timestep from floats in the range [-1, 1] writing only
- // num_features elements of input to (*this)[t], starting at offset.
- void NetworkIO::WriteTimeStepPart(int t, int offset, int num_features, const TFloat *input) {
-+ ASSERT_HOST(offset + num_features <= NumFeatures());
- if (int_mode_) {
- int8_t *line = i_[t] + offset;
- for (int i = 0; i < num_features; ++i) {
-diff --git a/unittest/lstm_layer_test.cc b/unittest/lstm_layer_test.cc
-new file mode 100644
-index 00000000..2c874995
---- /dev/null
-+++ b/unittest/lstm_layer_test.cc
-@@ -0,0 +1,177 @@
-+///////////////////////////////////////////////////////////////////////
-+// File: lstm_layer_test.cc
-+// Description: Tests that an NT_LSTM network layer with mutually
-+// inconsistent deserialized dimensions is rejected at
-+// load. The forward pass sizes its buffers from na_, no_
-+// and ns_ while the gate weight matrices drive their own
-+// dimensions, so a crafted .traineddata performs heap
-+// out-of-bounds writes/reads during the first
-+// recognition step (e.g. WriteTimeStepPart writing ns_
-+// floats into a source_ buffer sized from na_).
-+//
-+// Licensed under the Apache License, Version 2.0 (the "License");
-+// you may not use this file except in compliance with the License.
-+// You may obtain a copy of the License at
-+// http://www.apache.org/licenses/LICENSE-2.0
-+//
-+///////////////////////////////////////////////////////////////////////
-+
-+#include "include_gunit.h"
-+
-+#include "network.h" // for Network, NetworkType
-+#include "networkio.h"
-+#include "networkscratch.h"
-+#include "serialis.h" // for TFile
-+#include "stridemap.h"
-+
-+#include <cstdint>
-+#include <utility>
-+#include <vector>
-+
-+namespace tesseract {
-+namespace {
-+
-+// Appends raw little-endian values to a byte buffer.
-+class ByteWriter {
-+public:
-+ void PutU8(uint32_t v) { data_.push_back(static_cast<char>(v & 0xFF)); }
-+ void PutU32(uint32_t v) {
-+ for (int i = 0; i < 4; ++i) {
-+ data_.push_back(static_cast<char>((v >> (8 * i)) & 0xFF));
-+ }
-+ }
-+ void PutS32(int32_t v) { PutU32(static_cast<uint32_t>(v)); }
-+ const std::vector<char> &data() const { return data_; }
-+
-+private:
-+ std::vector<char> data_;
-+};
-+
-+void PutDoubleLE(ByteWriter *w, double d) {
-+ union {
-+ double d;
-+ uint64_t u;
-+ } conv;
-+ conv.d = d;
-+ w->PutU32(static_cast<uint32_t>(conv.u & 0xFFFFFFFF));
-+ w->PutU32(static_cast<uint32_t>(conv.u >> 32));
-+}
-+
-+// A serialized float-mode WeightMatrix with the given dimensions,
-+// all weight data zeroed.
-+void PutGateMatrix(ByteWriter *w, int32_t dim1, int32_t dim2) {
-+ w->PutU8(128); // mode: kDoubleFlag, float mode
-+ w->PutS32(dim1);
-+ w->PutS32(dim2);
-+ PutDoubleLE(w, 0.0); // empty_ cell
-+ for (int32_t i = 0; i < dim1 * dim2; ++i) {
-+ PutDoubleLE(w, 0.0);
-+ }
-+}
-+
-+// A serialized 1-D NT_LSTM network: header with the given ni/no, na_,
-+// then the four gates CI, GI, GF1, GO (GFS is not serialized for 1-D).
-+std::vector<char> MakeLstmNetwork(int ni, int no, int32_t na,
-+ const int32_t gate_dim1[4], const int32_t gate_dim2[4]) {
-+ ByteWriter w;
-+ w.PutU8(static_cast<uint32_t>(NT_LSTM));
-+ w.PutU8(0); // training: TS_DISABLED
-+ w.PutU8(0); // needs_to_backprop
-+ w.PutU32(0); // network_flags
-+ w.PutU32(static_cast<uint32_t>(ni));
-+ w.PutU32(static_cast<uint32_t>(no));
-+ w.PutU32(0); // num_weights
-+ w.PutU32(0); // name (empty string)
-+ w.PutS32(na);
-+ for (int g = 0; g < 4; ++g) {
-+ PutGateMatrix(&w, gate_dim1[g], gate_dim2[g]);
-+ }
-+ return w.data();
-+}
-+
-+// Builds the input a standalone LSTM layer would receive: one row of
-+// the given width with ni features.
-+NetworkIO MakeInput(int ni, int width) {
-+ StrideMap stride_map;
-+ stride_map.SetStride({{1, width}});
-+ NetworkIO input;
-+ input.ResizeToMap(false, stride_map, ni);
-+ return input;
-+}
-+
-+// Runs Forward on the loaded network; on unpatched code the out-of-
-+// bounds access this regression test guards against fires here.
-+void RunForward(Network *net, int ni, int width) {
-+ NetworkIO input = MakeInput(ni, width);
-+ NetworkScratch scratch;
-+ NetworkIO output;
-+ net->Forward(false, input, nullptr, &scratch, &output);
-+ delete net;
-+}
-+
-+// na_ must equal ni_ + nf_ + ns_ for a 1-D LSTM; here na_=2 but the
-+// CI matrix makes ns_=64, so the layer must be rejected. On unpatched
-+// code Forward writes 64 floats at offset ni_=1 into a source_ buffer
-+// sized for na_=2 (heap out-of-bounds write).
-+TEST(LstmLayerTest, RejectsInconsistentNa) {
-+ const int32_t dim1[4] = {64, 64, 64, 64};
-+ const int32_t dim2[4] = {3, 3, 3, 3};
-+ std::vector<char> bytes = MakeLstmNetwork(1, 1, 2, dim1, dim2);
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ Network *net = Network::CreateFromFile(&fp);
-+ if (net == nullptr) {
-+ return; // Fixed: the inconsistent layer is rejected at load.
-+ }
-+ RunForward(net, 1, 2);
-+ FAIL() << "crafted LSTM layer with inconsistent na_ was accepted";
-+}
-+
-+// All gate matrices must have dim1 == ns_; here the GI matrix has
-+// dim1=9 while ns_=5. On unpatched code the GI gate dot product writes
-+// 9 results into a temp line sized for 5 (heap out-of-bounds write).
-+TEST(LstmLayerTest, RejectsGateDim1Mismatch) {
-+ const int32_t dim1[4] = {5, 9, 5, 5};
-+ const int32_t dim2[4] = {7, 7, 7, 7};
-+ std::vector<char> bytes = MakeLstmNetwork(1, 5, 6, dim1, dim2);
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ Network *net = Network::CreateFromFile(&fp);
-+ if (net == nullptr) {
-+ return; // Fixed: the inconsistent layer is rejected at load.
-+ }
-+ RunForward(net, 1, 2);
-+ FAIL() << "crafted LSTM layer with inconsistent gate dim1 was accepted";
-+}
-+
-+// All gate matrices must have dim2 == na_ + 1; here the GI matrix has
-+// dim2=9 while na_=6. On unpatched code the GI gate dot product reads
-+// 8 inputs from a buffer sized for 6 (heap out-of-bounds read).
-+TEST(LstmLayerTest, RejectsGateDim2Mismatch) {
-+ const int32_t dim1[4] = {5, 5, 5, 5};
-+ const int32_t dim2[4] = {7, 9, 7, 7};
-+ std::vector<char> bytes = MakeLstmNetwork(1, 5, 6, dim1, dim2);
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ Network *net = Network::CreateFromFile(&fp);
-+ if (net == nullptr) {
-+ return; // Fixed: the inconsistent layer is rejected at load.
-+ }
-+ RunForward(net, 1, 2);
-+ FAIL() << "crafted LSTM layer with inconsistent gate dim2 was accepted";
-+}
-+
-+// A fully consistent 1-D LSTM layer must still be accepted and usable.
-+TEST(LstmLayerTest, AcceptsConsistentLayer) {
-+ const int32_t dim1[4] = {5, 5, 5, 5};
-+ const int32_t dim2[4] = {7, 7, 7, 7};
-+ std::vector<char> bytes = MakeLstmNetwork(1, 5, 6, dim1, dim2);
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ Network *net = Network::CreateFromFile(&fp);
-+ ASSERT_NE(net, nullptr);
-+ RunForward(net, 1, 2);
-+}
-+
-+} // namespace
-+} // namespace tesseract
deleted file mode 100644
@@ -1,209 +0,0 @@
-From 5426c23fe7fe32b62bd5694f68f71772f273c821 Mon Sep 17 00:00:00 2001
-From: Stefan Weil <sw@weilnetz.de>
-Date: Fri, 21 Aug 2026 16:56:51 +0200
-Subject: [PATCH] Reject recoder code values outside the sane range at load
-
-RecodedCharID::DeSerialize (hardened by 82727cc to check length_
-only) still read the individual code values as raw signed int32.
-UnicharCompress::ComputeCodeRange computes code_range_ as 1 plus the
-maximum code using a signed > comparison, so a code value of -1
-never raises the maximum and yields code_range_ = 0. SetupDecoder
-then resizes is_valid_start_ to 0 and writes is_valid_start_[code(0)]
-on the size-0 vector<bool>, an out-of-bounds write at a wild wrapped
-index (deterministic crash) on LSTMRecognizer load. A code value of
-INT32_MAX instead wraps code_range_ negative and makes resize()
-throw.
-
-Key changes:
-- unicharcompress.h: validate each deserialized code value to be
- within [0, UINT16_MAX), the same arbitrary cap used elsewhere for
- .traineddata counts; reject the recoder otherwise.
-- unicharcompress.h/.cpp: add defense-in-depth bounds assertions to
- IsValidFirstCode and SetupDecoder, matching the style of the
- NetworkIO assertions from 2f4d2f4.
-- unittest: add recoder_test, which feeds a crafted UnicharCompress
- with a -1 code and an INT32_MAX code and expects DeSerialize to
- fail (on unpatched code the -1 case dies on the SEGV in
- SetupDecoder, the huge case on the uncaught length_error), plus a
- positive control that valid codes load and answer
- code_range()/IsValidFirstCode() correctly.
-
-Reported-by: Zhixi "Jace" Sun <g.mygenie@gmail.com>
-Assisted-by: OpenCode / qwen3.8-27b-thinking (Alibaba Cloud)
-Signed-off-by: Stefan Weil <sw@weilnetz.de>
-(cherry picked from commit c94a5532ee04db5a4919542832fd94caee5ea58f)
-
-CVE: CVE-2026-88050
-Upstream-Status: Backport [https://github.com/tesseract-ocr/tesseract/commit/c94a5532ee04db5a4919542832fd94caee5ea58f]
-Signed-off-by: Ankur Tyagi <ankur.tyagi85@gmail.com>
----
- Makefile.am | 5 ++
- src/ccutil/unicharcompress.cpp | 1 +
- src/ccutil/unicharcompress.h | 13 ++++-
- unittest/recoder_test.cc | 91 ++++++++++++++++++++++++++++++++++
- 4 files changed, 109 insertions(+), 1 deletion(-)
- create mode 100644 unittest/recoder_test.cc
-
-diff --git a/Makefile.am b/Makefile.am
-index 4bcc858b..63ca6686 100644
---- a/Makefile.am
-+++ b/Makefile.am
-@@ -1202,6 +1202,7 @@ endif # !DISABLED_LEGACY_ENGINE
- check_PROGRAMS += progress_test
- check_PROGRAMS += qrsequence_test
- check_PROGRAMS += recodebeam_test
-+check_PROGRAMS += recoder_test
- check_PROGRAMS += rect_test
- check_PROGRAMS += resultiterator_test
- check_PROGRAMS += scanutils_test
-@@ -1435,6 +1436,10 @@ recodebeam_test_SOURCES = unittest/recodebeam_test.cc
- recodebeam_test_CPPFLAGS = $(unittest_CPPFLAGS)
- recodebeam_test_LDADD = $(TRAINING_LIBS) $(ICU_I18N_LIBS) $(ICU_UC_LIBS)
-
-+recoder_test_SOURCES = unittest/recoder_test.cc
-+recoder_test_CPPFLAGS = $(unittest_CPPFLAGS)
-+recoder_test_LDADD = $(TESS_LIBS)
-+
- rect_test_SOURCES = unittest/rect_test.cc
- rect_test_CPPFLAGS = $(unittest_CPPFLAGS)
- rect_test_LDADD = $(TESS_LIBS)
-diff --git a/src/ccutil/unicharcompress.cpp b/src/ccutil/unicharcompress.cpp
-index f909f76e..b3dd440f 100644
---- a/src/ccutil/unicharcompress.cpp
-+++ b/src/ccutil/unicharcompress.cpp
-@@ -400,6 +400,7 @@ void UnicharCompress::SetupDecoder() {
- for (unsigned c = 0; c < encoder_.size(); ++c) {
- const RecodedCharID &code = encoder_[c];
- decoder_[code] = c;
-+ ASSERT_HOST(code(0) >= 0 && code(0) < code_range_);
- is_valid_start_[code(0)] = true;
- RecodedCharID prefix = code;
- uint32_t len = code.length();
-diff --git a/src/ccutil/unicharcompress.h b/src/ccutil/unicharcompress.h
-index 9b696d03..6919b7d2 100644
---- a/src/ccutil/unicharcompress.h
-+++ b/src/ccutil/unicharcompress.h
-@@ -84,7 +84,17 @@ public:
- if (length_ > kMaxCodeLen) {
- return false;
- }
-- return fp->DeSerialize(&code_[0], length_);
-+ if (!fp->DeSerialize(&code_[0], length_)) {
-+ return false;
-+ }
-+ // Code values index arrays sized from the maximum code; reject values
-+ // that are out of the sane range for a recoded alphabet.
-+ for (uint32_t i = 0; i < length_; ++i) {
-+ if (code_[i] < 0 || code_[i] >= static_cast<int32_t>(UINT16_MAX)) {
-+ return false;
-+ }
-+ }
-+ return true;
- }
- bool operator==(const RecodedCharID &other) const {
- if (length_ != other.length_) {
-@@ -190,6 +200,7 @@ public:
- int DecodeUnichar(const RecodedCharID &code) const;
- // Returns true if the given code is a valid start or single code.
- bool IsValidFirstCode(int code) const {
-+ ASSERT_HOST(code >= 0 && code < code_range_);
- return is_valid_start_[code];
- }
- // Returns a list of valid non-final next codes for a given prefix code,
-diff --git a/unittest/recoder_test.cc b/unittest/recoder_test.cc
-new file mode 100644
-index 00000000..48dfa7ea
---- /dev/null
-+++ b/unittest/recoder_test.cc
-@@ -0,0 +1,91 @@
-+///////////////////////////////////////////////////////////////////////
-+// File: recoder_test.cc
-+// Description: Tests that a UnicharCompress (LSTM recoder) with code
-+// values outside the sane range is rejected at load.
-+// Negative code values leave code_range_ at zero, so
-+// SetupDecoder writes is_valid_start_[code(0)] out of
-+// bounds on a size-0 vector<bool>; huge code values wrap
-+// code_range_ and make resize() throw.
-+//
-+// Licensed under the Apache License, Version 2.0 (the "License");
-+// you may not use this file except in compliance with the License.
-+// You may obtain a copy of the License at
-+// http://www.apache.org/licenses/LICENSE-2.0
-+//
-+///////////////////////////////////////////////////////////////////////
-+
-+#include "include_gunit.h"
-+
-+#include "serialis.h" // for TFile
-+#include "unicharcompress.h"
-+
-+#include <cstdint>
-+#include <vector>
-+
-+namespace tesseract {
-+namespace {
-+
-+// Appends raw little-endian values to a byte buffer.
-+class ByteWriter {
-+public:
-+ void PutU8(uint32_t v) { data_.push_back(static_cast<char>(v & 0xFF)); }
-+ void PutU32(uint32_t v) {
-+ for (int i = 0; i < 4; ++i) {
-+ data_.push_back(static_cast<char>((v >> (8 * i)) & 0xFF));
-+ }
-+ }
-+ void PutS32(int32_t v) { PutU32(static_cast<uint32_t>(v)); }
-+ const std::vector<char> &data() const { return data_; }
-+
-+private:
-+ std::vector<char> data_;
-+};
-+
-+// A serialized UnicharCompress with one length-1 RecodedCharID per
-+// given code value (self-normalizing).
-+std::vector<char> MakeRecoder(const std::vector<int32_t> &codes) {
-+ ByteWriter w;
-+ w.PutU32(codes.size());
-+ for (int32_t code : codes) {
-+ w.PutU8(1); // self_normalized_
-+ w.PutU32(1); // length_
-+ w.PutS32(code); // code_[0]
-+ }
-+ return w.data();
-+}
-+
-+// A recoder code of -1 keeps code_range_ at 0, so on unpatched code
-+// SetupDecoder performs an out-of-bounds write into the size-0
-+// is_valid_start_ vector<bool>.
-+TEST(RecoderTest, RejectsNegativeCode) {
-+ std::vector<char> bytes = MakeRecoder({-1});
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ UnicharCompress recoder;
-+ EXPECT_FALSE(recoder.DeSerialize(&fp));
-+}
-+
-+// A recoder code of INT32_MAX wraps code_range_ to a negative value,
-+// so on unpatched code SetupDecoder's resize() throws.
-+TEST(RecoderTest, RejectsHugeCode) {
-+ std::vector<char> bytes = MakeRecoder({INT32_MAX});
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ UnicharCompress recoder;
-+ EXPECT_FALSE(recoder.DeSerialize(&fp));
-+}
-+
-+// A valid recoder must still be accepted and usable.
-+TEST(RecoderTest, AcceptsValidCodes) {
-+ std::vector<char> bytes = MakeRecoder({0, 1});
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ UnicharCompress recoder;
-+ ASSERT_TRUE(recoder.DeSerialize(&fp));
-+ EXPECT_EQ(recoder.code_range(), 2);
-+ EXPECT_TRUE(recoder.IsValidFirstCode(0));
-+ EXPECT_TRUE(recoder.IsValidFirstCode(1));
-+}
-+
-+} // namespace
-+} // namespace tesseract
deleted file mode 100644
@@ -1,184 +0,0 @@
-From 79bf602b0cd5216c125f86eaf3c768daae256930 Mon Sep 17 00:00:00 2001
-From: Stefan Weil <sw@weilnetz.de>
-Date: Fri, 21 Aug 2026 20:29:37 +0200
-Subject: [PATCH] Validate vector counts in GenericVector::read
-
-The callback form of GenericVector::read read two independent int32
-fields from the file: reserved sized the allocation via reserve(),
-while size_used_ drove the element loop. Neither was capped and no
-size_used_ <= reserved invariant was checked, so a crafted
-.traineddata (e.g. the fontinfo table of a version >= 4 inttemp
-component) performed a heap out-of-bounds write during legacy
-engine initialization.
-
-Key changes:
-- genericvector.h: reject negative or over-limit reserved
- (matching the 50000000 cap of the DeSerialize overloads) and
- reject size_used_ < 0 or size_used_ > reserved before entering
- the read loop. Legit files always satisfy size_used_ <= reserved,
- as write() persists size_reserved_ first.
-- unittest: add genericvector_test covering size_used_ beyond
- reserved (on unpatched code the ASan build dies on a
- heap-buffer-overflow in the read loop), negative counts, and a
- consistent vector that must still load.
-
-Reported-by: Zhixi "Jace" Sun <g.mygenie@gmail.com>
-Assisted-by: OpenCode / qwen3.8-27b-thinking (Alibaba Cloud)
-Signed-off-by: Stefan Weil <sw@weilnetz.de>
-(cherry picked from commit 56e09ca12e751623fe796ce1554ce704bffd2ef0)
-
-CVE: CVE-2026-88051
-Upstream-Status: Backport [https://github.com/tesseract-ocr/tesseract/commit/56e09ca12e751623fe796ce1554ce704bffd2ef0]
-Signed-off-by: Ankur Tyagi <ankur.tyagi85@gmail.com>
----
- Makefile.am | 5 ++
- src/ccutil/genericvector.h | 10 ++++
- unittest/genericvector_test.cc | 91 ++++++++++++++++++++++++++++++++++
- 3 files changed, 106 insertions(+)
- create mode 100644 unittest/genericvector_test.cc
-
-diff --git a/Makefile.am b/Makefile.am
-index cc748455..c67f0f63 100644
---- a/Makefile.am
-+++ b/Makefile.am
-@@ -1159,6 +1159,7 @@ check_PROGRAMS += equationdetect_test
- endif # !DISABLED_LEGACY_ENGINE
- check_PROGRAMS += fileio_test
- check_PROGRAMS += fullyconnected_test
-+check_PROGRAMS += genericvector_test
- check_PROGRAMS += heap_test
- check_PROGRAMS += imagedata_test
- if !DISABLED_LEGACY_ENGINE
-@@ -1298,6 +1299,10 @@ fullyconnected_test_SOURCES = unittest/fullyconnected_test.cc
- fullyconnected_test_CPPFLAGS = $(unittest_CPPFLAGS)
- fullyconnected_test_LDADD = $(TESS_LIBS)
-
-+genericvector_test_SOURCES = unittest/genericvector_test.cc
-+genericvector_test_CPPFLAGS = $(unittest_CPPFLAGS)
-+genericvector_test_LDADD = $(TESS_LIBS)
-+
- heap_test_SOURCES = unittest/heap_test.cc
- heap_test_CPPFLAGS = $(unittest_CPPFLAGS)
- heap_test_LDADD = $(TESS_LIBS)
-diff --git a/src/ccutil/genericvector.h b/src/ccutil/genericvector.h
-index 4a5bbe12..cbe1e203 100644
---- a/src/ccutil/genericvector.h
-+++ b/src/ccutil/genericvector.h
-@@ -654,10 +654,20 @@ bool GenericVector<T>::read(TFile *f, const std::function<bool(TFile *, T *)> &c
- if (f->FReadEndian(&reserved, sizeof(reserved), 1) != 1) {
- return false;
- }
-+ // Arbitrarily limit the number of elements to protect against bad data.
-+ const uint32_t limit = 50000000;
-+ if (reserved < 0 || static_cast<uint32_t>(reserved) > limit) {
-+ return false;
-+ }
- reserve(reserved);
- if (f->FReadEndian(&size_used_, sizeof(size_used_), 1) != 1) {
- return false;
- }
-+ // size_used_ is an independent file field; without this check the reads
-+ // below land past the end of the buffer sized from reserved.
-+ if (size_used_ < 0 || size_used_ > reserved) {
-+ return false;
-+ }
- if (cb != nullptr) {
- for (int i = 0; i < size_used_; ++i) {
- if (!cb(f, data_ + i)) {
-diff --git a/unittest/genericvector_test.cc b/unittest/genericvector_test.cc
-new file mode 100644
-index 00000000..269b00f0
---- /dev/null
-+++ b/unittest/genericvector_test.cc
-@@ -0,0 +1,91 @@
-+///////////////////////////////////////////////////////////////////////
-+// File: genericvector_test.cc
-+// Description: Tests that the callback form of GenericVector::read
-+// rejects vectors whose size_used_ exceeds reserved (or
-+// whose counts are negative). reserved sizes the buffer
-+// while size_used_ is an independent file field driving
-+// the element loop, so a crafted .traineddata (e.g. the
-+// fontinfo table of a version >= 4 inttemp component)
-+// performs a heap out-of-bounds write during legacy
-+// engine initialization.
-+//
-+// Licensed under the Apache License, Version 2.0 (the "License");
-+// you may not use this file except in compliance with the License.
-+// You may obtain a copy of the License at
-+// http://www.apache.org/licenses/LICENSE-2.0
-+//
-+///////////////////////////////////////////////////////////////////////
-+
-+#include "include_gunit.h"
-+
-+#include "genericvector.h"
-+#include "serialis.h" // for TFile
-+
-+#include <cstdint>
-+#include <vector>
-+
-+namespace tesseract {
-+namespace {
-+
-+// Appends raw little-endian values to a byte buffer.
-+class ByteWriter {
-+public:
-+ void PutS32(int32_t v) {
-+ uint32_t u = static_cast<uint32_t>(v);
-+ for (int i = 0; i < 4; ++i) {
-+ data_.push_back(static_cast<char>((u >> (8 * i)) & 0xFF));
-+ }
-+ }
-+ const std::vector<char> &data() const { return data_; }
-+
-+private:
-+ std::vector<char> data_;
-+};
-+
-+// A serialized vector header (reserved, size_used_) followed by the
-+// given number of int32 elements.
-+std::vector<char> MakeVector(int32_t reserved, int32_t size_used, int32_t num_elements) {
-+ ByteWriter w;
-+ w.PutS32(reserved);
-+ w.PutS32(size_used);
-+ for (int32_t i = 0; i < num_elements; ++i) {
-+ w.PutS32(i);
-+ }
-+ return w.data();
-+}
-+
-+// reserved=4 but size_used_=0x10000: on unpatched code the callback
-+// loop writes 65536 ints past the 4-int buffer (heap out-of-bounds
-+// write).
-+TEST(GenericVectorTest, RejectsSizeUsedBeyondReserved) {
-+ std::vector<char> bytes = MakeVector(4, 0x10000, 0x10000);
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ GenericVector<int> v;
-+ EXPECT_FALSE(v.read(&fp, [](TFile *f, int *p) { return f->DeSerialize(p); }));
-+}
-+
-+// Negative counts must be rejected; on unpatched code the read
-+// "succeeds" and leaves size_used_ negative.
-+TEST(GenericVectorTest, RejectsNegativeCounts) {
-+ std::vector<char> bytes = MakeVector(-1, -1, 0);
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ GenericVector<int> v;
-+ EXPECT_FALSE(v.read(&fp, [](TFile *f, int *p) { return f->DeSerialize(p); }));
-+}
-+
-+// A consistent vector must still be accepted.
-+TEST(GenericVectorTest, AcceptsConsistentVector) {
-+ std::vector<char> bytes = MakeVector(4, 2, 2);
-+ TFile fp;
-+ ASSERT_TRUE(fp.Open(bytes.data(), bytes.size()));
-+ GenericVector<int> v;
-+ ASSERT_TRUE(v.read(&fp, [](TFile *f, int *p) { return f->DeSerialize(p); }));
-+ EXPECT_EQ(v.size(), 2);
-+ EXPECT_EQ(v[0], 0);
-+ EXPECT_EQ(v[1], 1);
-+}
-+
-+} // namespace
-+} // namespace tesseract
deleted file mode 100644
@@ -1,165 +0,0 @@
-From c60ad589ff3a645a3fdf13cb24fe719313145a06 Mon Sep 17 00:00:00 2001
-From: Stefan Weil <sw@weilnetz.de>
-Date: Fri, 21 Aug 2026 18:34:33 +0200
-Subject: [PATCH] Reject unicharset files whose inserts desync id from unichars
-
-UNICHARSET::load_via_fgets reads the unichar count via sscanf and
-trusts it as the loop bound, indexing the unichars vector with the
-loop index id via the unchecked set_* accessors. unichar_insert is a
-no-op for duplicate (or empty) representations, so once any insert
-no-ops, unichars.size() falls behind id and the subsequent
-set_*(id, ...) and unichars[id].properties writes land past the end
-of the vector - a deterministic heap out-of-bounds write (including a
-std::string assignment via set_normed) for every remaining line, on
-both the LSTM and legacy init paths. A malformed unicharset with a
-duplicate line (e.g. two identical entries) triggers it; a
-non-positive header count likewise loads an empty unicharset
-"successfully".
-
-Key changes:
-- unicharset.cpp: reject unicharset_size <= 0, and after each insert
- verify the vector actually grew to id + 1; on mismatch report the
- offending line and reject the file instead of writing out of bounds.
-- unittest: add unicharset_load_test with a duplicate-representation
- unicharset (on unpatched code the test dies on the
- container-overflow in load_via_fgets), a zero and a negative count,
- and a positive control that a valid unicharset still loads.
-
-Reported-by: Zhixi "Jace" Sun <g.mygenie@gmail.com>
-Assisted-by: OpenCode / qwen3.8-27b-thinking (Alibaba Cloud)
-Signed-off-by: Stefan Weil <sw@weilnetz.de>
-(cherry picked from commit 2d04d640db2e8c7e3bab2369d599343b5a8b8443)
-
-CVE: CVE-2026-88052
-Upstream-Status: Backport [https://github.com/tesseract-ocr/tesseract/commit/2d04d640db2e8c7e3bab2369d599343b5a8b8443]
-Signed-off-by: Ankur Tyagi <ankur.tyagi85@gmail.com>
----
- Makefile.am | 5 +++
- src/ccutil/unicharset.cpp | 12 ++++++
- unittest/unicharset_load_test.cc | 64 ++++++++++++++++++++++++++++++++
- 3 files changed, 81 insertions(+)
- create mode 100644 unittest/unicharset_load_test.cc
-
-diff --git a/Makefile.am b/Makefile.am
-index bec9ab0e..22c2c7a0 100644
---- a/Makefile.am
-+++ b/Makefile.am
-@@ -1220,6 +1220,7 @@ check_PROGRAMS += tfile_test
- if ENABLE_TRAINING
- check_PROGRAMS += unichar_test
- check_PROGRAMS += unicharcompress_test
-+check_PROGRAMS += unicharset_load_test
- check_PROGRAMS += unicharset_test
- check_PROGRAMS += validate_grapheme_test
- check_PROGRAMS += validate_indic_test
-@@ -1493,6 +1494,10 @@ unicharcompress_test_SOURCES = unittest/unicharcompress_test.cc
- unicharcompress_test_CPPFLAGS = $(unittest_CPPFLAGS)
- unicharcompress_test_LDADD = $(TRAINING_LIBS) $(ICU_UC_LIBS)
-
-+unicharset_load_test_SOURCES = unittest/unicharset_load_test.cc
-+unicharset_load_test_CPPFLAGS = $(unittest_CPPFLAGS)
-+unicharset_load_test_LDADD = $(TESS_LIBS)
-+
- unicharset_test_SOURCES = unittest/unicharset_test.cc
- unicharset_test_CPPFLAGS = $(unittest_CPPFLAGS)
- unicharset_test_LDADD = $(TRAINING_LIBS) $(ICU_UC_LIBS)
-diff --git a/src/ccutil/unicharset.cpp b/src/ccutil/unicharset.cpp
-index b29ec3b7..0e72ae48 100644
---- a/src/ccutil/unicharset.cpp
-+++ b/src/ccutil/unicharset.cpp
-@@ -791,6 +791,9 @@ bool UNICHARSET::load_via_fgets(
- sscanf(buffer, "%d", &unicharset_size) != 1) {
- return false;
- }
-+ if (unicharset_size <= 0) {
-+ return false;
-+ }
- for (UNICHAR_ID id = 0; id < unicharset_size; ++id) {
- char unichar[256];
- unsigned int properties;
-@@ -884,6 +887,15 @@ bool UNICHARSET::load_via_fgets(
- } else {
- this->unichar_insert_backwards_compatible(unichar);
- }
-+ // A duplicate or empty representation makes the insert a no-op,
-+ // desynchronizing id from the unichars vector; the set_* calls and
-+ // unichars[id] below would then write out of bounds. The file is
-+ // malformed, so reject it.
-+ if (size() != static_cast<size_t>(id) + 1) {
-+ fprintf(stderr, "%s:%d unichar %d has a duplicate or empty representation\n",
-+ __FILE__, __LINE__, id);
-+ return false;
-+ }
-
- this->set_isalpha(id, properties & ISALPHA_MASK);
- this->set_islower(id, properties & ISLOWER_MASK);
-diff --git a/unittest/unicharset_load_test.cc b/unittest/unicharset_load_test.cc
-new file mode 100644
-index 00000000..d775e233
---- /dev/null
-+++ b/unittest/unicharset_load_test.cc
-@@ -0,0 +1,64 @@
-+///////////////////////////////////////////////////////////////////////
-+// File: unicharset_load_test.cc
-+// Description: Tests that UNICHARSET::load_via_fgets rejects unicharset
-+// files whose insertions desynchronize the id loop index
-+// from the unichars vector (duplicate or empty
-+// representations), which would make the subsequent set_*
-+// calls write out of bounds, and non-positive size counts.
-+//
-+// Licensed under the Apache License, Version 2.0 (the "License");
-+// you may not use this file except in compliance with the License.
-+// You may obtain a copy of the License at
-+// http://www.apache.org/licenses/LICENSE-2.0
-+//
-+///////////////////////////////////////////////////////////////////////
-+
-+#include "include_gunit.h"
-+
-+#include "serialis.h" // for TFile
-+#include "unicharset.h"
-+
-+#include <cstring>
-+
-+namespace tesseract {
-+namespace {
-+
-+// Loads the given unicharset text via the TFile-based loader.
-+bool LoadUnicharset(const char *text, UNICHARSET *unicharset) {
-+ TFile fp;
-+ if (!fp.Open(text, std::strlen(text))) {
-+ return false;
-+ }
-+ return unicharset->load_from_file(&fp, false);
-+}
-+
-+// A duplicate representation makes the second insert a no-op, so on
-+// unpatched code the set_* calls for the remaining lines write past
-+// the end of the unichars vector (ASan container-overflow).
-+TEST(UnicharsetLoadTest, RejectsDuplicateRepresentation) {
-+ const char *text = "3\nA 0 Latin\nA 0 Latin\nB 0 Latin\n";
-+ UNICHARSET unicharset;
-+ EXPECT_FALSE(LoadUnicharset(text, &unicharset));
-+}
-+
-+// A non-positive size count must be rejected; on unpatched code a
-+// zero or negative count loads an empty unicharset successfully.
-+TEST(UnicharsetLoadTest, RejectsNonPositiveCount) {
-+ const char *texts[] = {"0\n", "-1\n"};
-+ for (const char *text : texts) {
-+ UNICHARSET unicharset;
-+ EXPECT_FALSE(LoadUnicharset(text, &unicharset));
-+ }
-+}
-+
-+// A valid unicharset must still be accepted.
-+TEST(UnicharsetLoadTest, AcceptsValidUnicharset) {
-+ const char *text = "3\nA 0 Latin\nB 0 Latin\nC 0 Latin\n";
-+ UNICHARSET unicharset;
-+ ASSERT_TRUE(LoadUnicharset(text, &unicharset));
-+ EXPECT_EQ(unicharset.size(), 3u);
-+ EXPECT_STREQ(unicharset.id_to_unichar(1), "B");
-+}
-+
-+} // namespace
-+} // namespace tesseract
deleted file mode 100644
@@ -1,274 +0,0 @@
-From 83bbac0975f52a5171776fddcc800a2aaa5206a3 Mon Sep 17 00:00:00 2001
-From: Stefan Weil <sw@weilnetz.de>
-Date: Fri, 21 Aug 2026 14:38:55 +0200
-Subject: [PATCH] Reject empty network stacks in LSTM .traineddata
- deserialization
-
-Plumbing::DeSerialize read the network stack size from the untrusted
-TESSDATA_LSTM component and only rejected size > 10000. A crafted
-.traineddata with a top-level Series/Parallel/Reversed layer with an
-empty stack survived load; LSTMRecognizer initialization then called
-network_->CacheXScaleFactor(network_->XScaleFactor()), which
-dereferences stack_[0] on the empty vector:
-- Series: Series::CacheXScaleFactor (series.cpp)
-- Parallel/Reversed: inherited Plumbing::XScaleFactor
-The virtual call through the wild pointer crashed the process at
-initialization (deterministic denial of service).
-
-Key changes:
-- plumbing.cpp: reject size == 0 for all plumbing types, and
- size < 2 for NT_SERIES (Series::Forward requires two or more
- networks and always aborts on one).
-- tessedit.cpp: fail the language load gracefully when the LSTM model
- cannot be loaded, instead of aborting via ASSERT_HOST, so
- TessBaseAPI::Init returns -1 on corrupt traineddata.
-- unittest: add plumbing_test, which builds a minimal traineddata
- with empty/undersized LSTM plumbing stacks and expects a graceful
- init failure. On unpatched code the tests die on the original
- SEGV in Series::CacheXScaleFactor / Plumbing::XScaleFactor.
-
-Reported-by: Zhixi "Jace" Sun <g.mygenie@gmail.com>
-Assisted-by: OpenCode / qwen3.8-27b-thinking (Alibaba Cloud)
-Signed-off-by: Stefan Weil <sw@weilnetz.de>
-(cherry picked from commit 552771236b0d80cbdb0c7dd856120fa21a4672e5)
-
-CVE: CVE-2026-88054
-Upstream-Status: Backport [https://github.com/tesseract-ocr/tesseract/commit/552771236b0d80cbdb0c7dd856120fa21a4672e5]
-Signed-off-by: Ankur Tyagi <ankur.tyagi85@gmail.com>
----
- Makefile.am | 5 ++
- src/ccmain/tessedit.cpp | 7 +-
- src/lstm/plumbing.cpp | 6 ++
- unittest/plumbing_test.cc | 165 ++++++++++++++++++++++++++++++++++++++
- 4 files changed, 182 insertions(+), 1 deletion(-)
- create mode 100644 unittest/plumbing_test.cc
-
-diff --git a/Makefile.am b/Makefile.am
-index c67f0f63..a6ee0931 100644
---- a/Makefile.am
-+++ b/Makefile.am
-@@ -1198,6 +1198,7 @@ if ENABLE_TRAINING
- check_PROGRAMS += pango_font_info_test
- endif # ENABLE_TRAINING
- check_PROGRAMS += paragraphs_test
-+check_PROGRAMS += plumbing_test
- if !DISABLED_LEGACY_ENGINE
- check_PROGRAMS += params_model_test
- endif # !DISABLED_LEGACY_ENGINE
-@@ -1427,6 +1428,10 @@ paragraphs_test_SOURCES = unittest/paragraphs_test.cc
- paragraphs_test_CPPFLAGS = $(unittest_CPPFLAGS)
- paragraphs_test_LDADD = $(TESS_LIBS)
-
-+plumbing_test_SOURCES = unittest/plumbing_test.cc
-+plumbing_test_CPPFLAGS = $(unittest_CPPFLAGS)
-+plumbing_test_LDADD = $(TESS_LIBS)
-+
- if !DISABLED_LEGACY_ENGINE
- params_model_test_SOURCES = unittest/params_model_test.cc
- params_model_test_CPPFLAGS = $(unittest_CPPFLAGS)
-diff --git a/src/ccmain/tessedit.cpp b/src/ccmain/tessedit.cpp
-index 7ec89bd9..722ed09b 100644
---- a/src/ccmain/tessedit.cpp
-+++ b/src/ccmain/tessedit.cpp
-@@ -169,7 +169,12 @@ bool Tesseract::init_tesseract_lang_data(const std::string &language, OcrEngineM
- #endif // ndef DISABLED_LEGACY_ENGINE
- if (mgr->IsComponentAvailable(TESSDATA_LSTM)) {
- lstm_recognizer_ = new LSTMRecognizer(language_data_path_prefix.c_str());
-- ASSERT_HOST(lstm_recognizer_->Load(this->params(), lstm_use_matrix ? language : "", mgr));
-+ if (!lstm_recognizer_->Load(this->params(), lstm_use_matrix ? language : "", mgr)) {
-+ delete lstm_recognizer_;
-+ lstm_recognizer_ = nullptr;
-+ tprintf("Error: Failed to load the LSTM model from %s\n", tessdata_path.c_str());
-+ return false;
-+ }
- } else {
- #ifdef DISABLED_LEGACY_ENGINE
- // The legacy engine is compiled out, so we cannot fall back to it.
-diff --git a/src/lstm/plumbing.cpp b/src/lstm/plumbing.cpp
-index f0133148..15cf3d62 100644
---- a/src/lstm/plumbing.cpp
-+++ b/src/lstm/plumbing.cpp
-@@ -226,6 +226,12 @@ bool Plumbing::DeSerialize(TFile *fp) {
- if (size > 10000) {
- return false;
- }
-+ // Reject empty stacks: XScaleFactor, CacheXScaleFactor and other methods
-+ // unconditionally dereference stack_[0] during network initialization.
-+ // A Series needs at least two networks (see Series::Forward).
-+ if (size == 0 || (type() == NT_SERIES && size == 1)) {
-+ return false;
-+ }
- for (uint32_t i = 0; i < size; ++i) {
- Network *network = CreateFromFile(fp);
- if (network == nullptr) {
-diff --git a/unittest/plumbing_test.cc b/unittest/plumbing_test.cc
-new file mode 100644
-index 00000000..27486f86
---- /dev/null
-+++ b/unittest/plumbing_test.cc
-@@ -0,0 +1,165 @@
-+///////////////////////////////////////////////////////////////////////
-+// File: plumbing_test.cc
-+// Description: Tests that a corrupt TESSDATA_LSTM component in a
-+// .traineddata file is rejected without crashing. A
-+// plumbing layer (Series/Parallel/Reversed) with an
-+// empty or undersized network stack would make
-+// XScaleFactor/CacheXScaleFactor dereference stack_[0]
-+// during engine initialization.
-+//
-+// Licensed under the Apache License, Version 2.0 (the "License");
-+// you may not use this file except in compliance with the License.
-+// You may obtain a copy of the License at
-+// http://www.apache.org/licenses/LICENSE-2.0
-+//
-+///////////////////////////////////////////////////////////////////////
-+
-+#include "include_gunit.h"
-+
-+#include <tesseract/baseapi.h>
-+
-+#include "network.h" // for NetworkType
-+#include "tessdatamanager.h" // for TessdataManager, TESSDATA_LSTM
-+
-+#include <cstdint>
-+#include <cstdio>
-+#include <cstdlib>
-+#include <cstring>
-+#include <string>
-+#include <vector>
-+
-+namespace tesseract {
-+namespace {
-+
-+// Minimal unicharset with the special codes (space, Joined, Broken) that
-+// UNICHARSET::load_from_file expects, as embedded in the TESSDATA_LSTM
-+// component by LSTMRecognizer::Serialize.
-+const char kMinUnicharset[] =
-+ "3\n"
-+ "NULL 0 NULL 0\n"
-+ "Joined 7 0,69,188,255,486,1218,0,30,486,1188 Latin 26 0 98 Joined\n"
-+ "|Broken|0|1 f 0,69,186,255,892,2138,0,80,892,2058 Common 84 10 84 |Broken|0|1\n";
-+
-+// Appends raw little-endian values to a byte buffer.
-+class ByteWriter {
-+public:
-+ void PutU8(uint32_t v) { data_.push_back(static_cast<char>(v & 0xFF)); }
-+ void PutU32(uint32_t v) {
-+ for (int i = 0; i < 4; ++i) {
-+ data_.push_back(static_cast<char>((v >> (8 * i)) & 0xFF));
-+ }
-+ }
-+ void PutS32(int32_t v) { PutU32(static_cast<uint32_t>(v)); }
-+ void PutString(const char *s) {
-+ PutU32(static_cast<uint32_t>(std::strlen(s)));
-+ data_.insert(data_.end(), s, s + std::strlen(s));
-+ }
-+ void PutRaw(const char *s) { data_.insert(data_.end(), s, s + std::strlen(s)); }
-+ const std::vector<char> &data() const { return data_; }
-+
-+private:
-+ std::vector<char> data_;
-+};
-+
-+// Serialized network header as written by Network::Serialize:
-+// int8 type, int8 training, int8 needs_to_backprop, int32 network_flags,
-+// int32 ni, int32 no, int32 num_weights, string name.
-+void AppendNetworkHeader(ByteWriter *w, NetworkType type) {
-+ w->PutU8(static_cast<uint32_t>(type));
-+ w->PutU8(0); // training: TS_DISABLED
-+ w->PutU8(0); // needs_to_backprop
-+ w->PutU32(0); // network_flags
-+ w->PutU32(0); // ni
-+ w->PutU32(0); // no
-+ w->PutU32(0); // num_weights
-+ w->PutString(""); // name
-+}
-+
-+// A minimal valid child network (NT_INPUT with a 1x1x1x1 shape).
-+void AppendInputChild(ByteWriter *w) {
-+ AppendNetworkHeader(w, NT_INPUT);
-+ w->PutS32(1); // batch
-+ w->PutS32(1); // height
-+ w->PutS32(1); // width
-+ w->PutS32(1); // depth
-+ w->PutS32(0); // loss type
-+}
-+
-+// Builds a TESSDATA_LSTM component whose top-level network is a plumbing
-+// layer of the given type with the given (corrupt) stack size, followed by
-+// the remaining fields of LSTMRecognizer::DeSerialize.
-+std::vector<char> MakeLstmComponent(NetworkType type, uint32_t stack_size) {
-+ ByteWriter w;
-+ AppendNetworkHeader(&w, type);
-+ w.PutU32(stack_size); // Plumbing::DeSerialize reads this as uint32
-+ for (uint32_t i = 0; i < stack_size; ++i) {
-+ AppendInputChild(&w);
-+ }
-+ w.PutRaw(kMinUnicharset); // unicharset (raw text, no recoder/unicharset components)
-+ w.PutString(""); // network_str_
-+ w.PutS32(0); // training_flags_
-+ w.PutS32(0); // training_iteration_
-+ w.PutS32(0); // sample_iteration_
-+ w.PutS32(0); // null_char_
-+ w.PutU32(0); // adam_beta_ (float 0.0)
-+ w.PutU32(0); // learning_rate_ (float 0.0)
-+ w.PutU32(0); // momentum_ (float 0.0)
-+ return w.data();
-+}
-+
-+// Writes a traineddata file with the given (corrupt) LSTM component to
-+// dir/eng.traineddata.
-+bool WriteCorruptTraineddata(const std::string &dir, const std::vector<char> &lstm) {
-+ TessdataManager mgr;
-+ mgr.OverwriteEntry(TESSDATA_LSTM, lstm.data(), static_cast<int>(lstm.size()));
-+ return mgr.SaveFile((dir + "/eng.traineddata").c_str(), nullptr);
-+}
-+
-+class PlumbingTest : public testing::Test {
-+protected:
-+ void SetUp() override {
-+ tmpl_ = "/tmp/tess_plumbing_test_XXXXXX";
-+ char *dir = mkdtemp(tmpl_.data());
-+ ASSERT_NE(dir, nullptr);
-+ dir_ = dir;
-+ }
-+ void TearDown() override {
-+ std::remove((dir_ + "/eng.traineddata").c_str());
-+ rmdir(dir_.c_str());
-+ }
-+ // Expects the LSTM engine to reject the corrupted traineddata
-+ // gracefully (init failure) instead of crashing.
-+ void ExpectInitFails(const std::vector<char> &lstm) {
-+ ASSERT_TRUE(WriteCorruptTraineddata(dir_, lstm));
-+ tesseract::TessBaseAPI api;
-+ EXPECT_EQ(api.Init(dir_.c_str(), "eng", tesseract::OEM_LSTM_ONLY), -1);
-+ }
-+ std::string dir_;
-+ std::string tmpl_;
-+};
-+
-+// Empty NT_SERIES stack: Series::CacheXScaleFactor would dereference
-+// stack_[0] on the empty vector during initialization.
-+TEST_F(PlumbingTest, RejectsEmptySeriesStack) {
-+ ExpectInitFails(MakeLstmComponent(NT_SERIES, 0));
-+}
-+
-+// Empty NT_PARALLEL stack: Plumbing::XScaleFactor would dereference
-+// stack_[0] on the empty vector during initialization.
-+TEST_F(PlumbingTest, RejectsEmptyParallelStack) {
-+ ExpectInitFails(MakeLstmComponent(NT_PARALLEL, 0));
-+}
-+
-+// Empty NT_XREVERSED stack: same crash as the parallel case.
-+TEST_F(PlumbingTest, RejectsEmptyReversedStack) {
-+ ExpectInitFails(MakeLstmComponent(NT_XREVERSED, 0));
-+}
-+
-+// A Series with a single network: Series::Forward requires at least two
-+// networks, so such a model can never work.
-+TEST_F(PlumbingTest, RejectsSingleNetworkSeries) {
-+ ExpectInitFails(MakeLstmComponent(NT_SERIES, 1));
-+}
-+
-+} // namespace
-+} // namespace tesseract
similarity index 63%
rename from meta-oe/recipes-graphics/tesseract/tesseract_5.5.3.bb
rename to meta-oe/recipes-graphics/tesseract/tesseract_5.5.4.bb
@@ -5,15 +5,8 @@ BUGTRACKER = "https://github.com/tesseract-ocr/tesseract/issues"
LICENSE = "Apache-2.0"
LIC_FILES_CHKSUM = "file://LICENSE;md5=3b83ef96387f14655fc854ddc3c6bd57"
-SRCREV = "db0ec62f81b0737fbbe184d8fea40af5738f8eef"
+SRCREV = "7e548e6eff172f38f97664afe7f7ce40db36c8d6"
SRC_URI = "git://github.com/${BPN}-ocr/${BPN}.git;branch=main;protocol=https;tag=${PV} \
- file://CVE-2026-88052.patch \
- file://CVE-2026-88048.patch \
- file://CVE-2026-88049.patch \
- file://CVE-2026-88050.patch \
- file://CVE-2026-88047.patch \
- file://CVE-2026-88051.patch \
- file://CVE-2026-88054.patch \
"
DEPENDS = "leptonica"