github tesseract-ocr/tesseract 5.5.4

4 hours ago

What's Changed

This release fixes 17 security advisories and several smaller issues.
In addition, it provides some smaller updates and improvements:

  • fix build of training tool on windows with VS2022: by @zdenop in #4590
  • arch,viewer: add missing and by @tachalorah in #4592
  • Fix recursive static initialization in SIMDDetect::GetDetector() by @stweil with @Copilot in #4598
  • exports GetDetector() from the tesseract DLL, so linker can resolve it. by @zdenop in #4602
  • Fix two \param names that refer to the member, not the argument by @darkdi in #4594
  • Add new GitHub workflow which checks commits in a pull request by @stweil in #4604
  • Build unit tests without training tools by @stweil in #4596
  • docs: update dead Tesseract training docs link by @yzxcj797 in #4605
  • Fix several high and medium priority issues reported by Coverity Scan by @stweil in #4597
  • Fix Coverity warnings by @stweil in #4606
  • Avoid quadratic paragraph restart scan by @ilkecan in #4599
  • Bump actions/checkout from 4 to 7 by @dependabot[bot] in #4607
  • Fix GitHub AI findings from 'Security and quality' by @stweil in #4613
  • Fix inefficient empty string test in ScrollView menus by @stweil in #4612
  • Fix some issues reported by Coverity by @stweil in #4611
  • fix: prevent out-of-bounds read in UNICHAR::UTF8ToUTF32 by @SomSamantray in #4609
  • fix: prevent out-of-bounds access on truncated UTF-8 strings and crafted index maps by @stweil in #4617
  • Avoid redundant result iterator restarts by @ilkecan in #4601
  • fix: potential memory leak by @y1lan in #4622
  • Avoid float overflow for unclassified-word certainty by @stweil in #4630
  • Reject int-mode weight matrices with zero inputs in the deserializer by @stweil in #4644
  • Fix position-dependent paragraph bounding box by @stweil in #4632
  • Reject negative/oversized class ids in legacy inttemp by @stweil in #4640
  • Exit child process when execvp fails by @Ramya-9353 in #4577
  • Neutralise file-sourced pointers in ReadAdaptedClass before use by @stweil in #4645
  • Fix Devanagari font download URL by @SKSavale in #4654
  • cmake: allow unit tests to use installed GoogleTest by @manduinca in #4650
  • Reject malformed dimensions in GENERIC_2D_ARRAY deserialization by @stweil in #4649
  • ccstruct: fix leak of orphaned outline cycles in EliminateDuplicateOutlines by @stweil in #4655
  • Fix double-free/UAF in AppendLinePolygon for RTL PAGE XML output by @stweil in #4647
  • wordrec: bound seam search to prevent timeouts on degenerate input by @stweil in #4656
  • lstm: reject old-format WeightMatrix instead of aborting by @stweil in #4657
  • wordrec: add a per-pass seam search budget to bound many-word timeouts by @stweil in #4659
  • Reject out-of-range null_char in LSTM model load by @stweil in #4642
  • Harden class-id handling in the int/adaptive classifier by @stweil in #4639
  • Reject int-mode weight matrices with a short scale vector by @stweil in #4641
  • Reject a recoder whose code range exceeds the network output count by @stweil in #4643
  • Skip unit tests when their traineddata model is missing by @stweil in #4646
  • nsis: rewrite winpath helper without a fixed-size buffer by @stweil in #4661
  • Fix C++17 build of constexpr FCOORD operators by @stweil in #4665
  • Fix use-after-free in ReplaceCurrentWord with several languages by @mardgez in #4667

New Contributors

Full Changelog: 5.5.3...5.5.4

Don't miss a new tesseract release

NewReleases is sending notifications on new releases.