Fix uninitialized memory read by kintel · Pull Request #1054 · AngusJohnson/Clipper2

kintel · 2026-01-11T15:14:52Z

This is a clear uninitialized memory read.

Note that some compilers are able to detect this kind of memory access through static analysis when -Wall is enabled (specifically through -Werror=maybe-uninitialized). However, the capabilities of this features varies between compiler versions. gcc-11.4.0 in Ubuntu-22.04 seems to be particularly good at this, but this compiler is not used by Clipper's CI, which is why this issue went undetected.

* Bumped Clipper to 2.0.1 * Add bugfix for AngusJohnson/Clipper2#1054

* Run once daily. * Update (and run) translation scripts; require bash; enable error options. * Bumped Clipper to 2.0.1 (openscad#6494) * Bumped Clipper to 2.0.1 * Add bugfix for AngusJohnson/Clipper2#1054 --------- Co-authored-by: Torsten Paul <[email protected]> Co-authored-by: Marius Kintel <[email protected]>

* feat: more properties in __getitem__ * chore: cleanup * chore: add tests * chore: merge 91 commits from openscad/master (#277) * add correct StartupWMClass to set the correct icon on Gnome I believe this is the XDG way to do this, but I am not sure. * Check for save/discard/cancel on quit. Fixes #3971. (#6311) * Bump CGAL to 6.1.0 (#6318) * Make vector swizzle experimental (#6307) * Don't log to stderr (#6323) * Fully separate zoom levels for each tab. Fixes #4742. (#6320) * Fix/cleanup compiler warnings openscad/openscad#6245 (#6300) * Remove obsolete HTML example converter (#6328) * reformat python tests scripts in pep8 (#6330) * Bump the github-actions group with 3 updates Bumps the github-actions group with 3 updates: [actions/upload-artifact](https://github.com/actions/upload-artifact), [github/codeql-action](https://github.com/github/codeql-action) and [actions/stale](https://github.com/actions/stale). Updates `actions/upload-artifact` from 4 to 5 - [Release notes](https://github.com/actions/upload-artifact/releases) - [Commits](https://github.com/actions/upload-artifact/compare/v4...v5) Updates `github/codeql-action` from 3 to 4 - [Release notes](https://github.com/github/codeql-action/releases) - [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md) - [Commits](https://github.com/github/codeql-action/compare/v3...v4) Updates `actions/stale` from 9 to 10 - [Release notes](https://github.com/actions/stale/releases) - [Changelog](https://github.com/actions/stale/blob/main/CHANGELOG.md) - [Commits](https://github.com/actions/stale/compare/v9...v10) --- updated-dependencies: - dependency-name: actions/upload-artifact dependency-version: '5' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions - dependency-name: github/codeql-action dependency-version: '4' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions - dependency-name: actions/stale dependency-version: '10' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions ... Signed-off-by: dependabot[bot] <[email protected]> * Check print service name only if remote print service is selected. This fixes the issue that the print dialog is always shown if the local application is selected for printing. * Don't crash MouseSelector, debug GL init (#6334) * beautify (#6342) * Set QSCINTILLA_DLL for MINGW (and not MXECROSS). (#6343) * Warn on offset() with both r and delta specified (#6348) * Fix QFile::Open nodiscard warning with Qt6 6.10.0. (#6364) * Update cmdline help to show Manifold as default (#6350) * Rm unused function. Fixes #6352. (#6353) * Add Color module spec (#6339) * Enables "Ctrl/Cmd-Mouse-wheel zooms text" config option to work (Preliminary fix for #6368) * Entirely removed superfluous EditorInterface::wheelEvent method * Bump Qt to 6.5.7 (#6371) * CircleCI: Update macOS resources to 26.1 on M4 (#6324) * Refactor Discretization Special Usage (#6251) * Tessallation specials $fn$fs$fa now grouped in an object Instead of fn,fs,fa as three doubles, they are passed as an object packaged up with the methods on them. * Repair basic functionality (#6376) * Fix indentation of CC0 template text. (#6375) Also clean up line endings. Add templates/README.md * Fix a number of non-ASCII file name issues (#6136) * Try caching MSYS2 packages Fixes #6381 if it works * Fix `if` in yaml wrong * Don't want `if`s anyway doesn't matter I did them wrong * Update msys2 gh actions caching Make one new cache a day from scratch. Leave note why we don't find the most recent cache. Skip saving since caches are immutable. * Comment updates + build up cache over a month * We can read the determined msys2 directory We're not restoring before the setup, like I originally planned. * Updating script to output list of pacboy packages to workflow Easier than my other approach, actually. Ah, whatever. * Don't change $MSYSTEM in GH Actions * Oops, I need to specify the bash shell I think * Use arrays instead of backslash line termination * Don't fail fast because it fails due to caching a lot well, until we actually get one * std::wstring_convert and std::codecvt_utf8_utf16 is going to be depreciated. Use boost::nowide::narrow (#6356) * Add Unit tests to fix measuring (#6299) * Starting point of release build CI workflow (#6374) * Define executors separately * Added OpenSCAD release build job with create-tarball * Centralize version string management (#6388) * Centralize version string management * Identify build tag in OPENSCAD_VERSION, and correctly extract OPENSCAD_SHORTVERSION * cimg/python: no 'latest' tag. Use 3.14 (#6399) * Disable WIP macOS release build (#6398) * CircleCI Bugfix: Use OPENSCAD_VERSION instead of VERSION (#6400) * create-tarball: Amend files manually since git-archive-all doesn't consider unversioned files (#6401) * cd to /tmp/artifacts before creating shasums (#6402) * Add CMake support + shell helpers for VERSION.txt and COMMIT.txt (#6403) * Ingest VERSION.txt and COMMIT.txt at cmake time * Shell helper for establishing version from VERSION.txt and COMMIT.txt * Snapshot builds now uses the version helpers to unify version management, also deprecated 32-bit windows (#6404) * Bugfix: binary name on macOS needs correct case (#6409) * Parametrize macOS build for release vs. snapshot (#6406) * Parametrize macOS build for release vs. snapshot * Use establish_version.sh and don't require git * Always checkout in the same way; with submodules * Bump actions/checkout from 5 to 6 in the github-actions group (#6418) Bumps the github-actions group with 1 update: [actions/checkout](https://github.com/actions/checkout). Updates `actions/checkout` from 5 to 6 - [Release notes](https://github.com/actions/checkout/releases) - [Changelog](https://github.com/actions/checkout/blob/main/CHANGELOG.md) - [Commits](https://github.com/actions/checkout/compare/v5...v6) --- updated-dependencies: - dependency-name: actions/checkout dependency-version: '6' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * Delete discretizer keyword args PyArg_ParseTupleAndKeywords doesn't allow keyword args it doesn't know about, so we delete them if found. Manually testing the fix. * Fix formatting of "Browser" header (#6433) Markdown rendered is confused by non-breaking space (`U+A0`) - replaced with a regular space. * Make find highlighting case-insensitive * chore: codegen version into manpage (#6448) * chore: codegen version into manpage testing process: - created a VERSION.txt file - ran the CMake build - confirm that the output `build/doc/openscad.1` has the same version as VERSION.txt * remove extra comment * Added job parameters openscad_build_type and openscad_platform (#6449) * Remove deprecated child() and assign(). (#6428) * Remove deprecated child() and assign(). Fixes #5557. Fixes #6425. * Update releases/next.md: child(), assign(), object(), textmetrics(), fontmetrics Remove deprecated child() and assign(). Add object(). Reclassify textmetrics() and fontmetrics() as functions. * More CircleCI cleanup (#6450) * Bump manifold to 3.3.2 and onetbb to 2022.3.0 (#6455) * Added failing test for #6456 (#6459) * Bumped manifold to fix #6235 (#6467) * Temporarily restore exact job names used by circleci-download-artifacts.py (#6469) * CircleCI: Update name of wasm build (#6470) * CircleCI release builds (#6412) * Move version info to .c file (#6463) * chore: bump nix-shell for latest version * convert to flake (for lockfile) * use stable release * remove duplicate lock entry * Add warning to cmake output that ENABLE_CGAL=OFF is a dev-only flag. * Fix font sample rendering in font list window. * Remove obsolete FontListDialog. * Change console font size via mouse wheel event (fixes #6186). This is only changing the current console widget, not the value in the preferences. * Align naming of release jobs, correct naming of tarball (#6484) * Bugfix: Forgot to update dependencies on tarball (#6485) * Bump the github-actions group with 2 updates Bumps the github-actions group with 2 updates: [actions/upload-artifact](https://github.com/actions/upload-artifact) and [actions/cache](https://github.com/actions/cache). Updates `actions/upload-artifact` from 5 to 6 - [Release notes](https://github.com/actions/upload-artifact/releases) - [Commits](https://github.com/actions/upload-artifact/compare/v5...v6) Updates `actions/cache` from 4 to 5 - [Release notes](https://github.com/actions/cache/releases) - [Changelog](https://github.com/actions/cache/blob/main/RELEASES.md) - [Commits](https://github.com/actions/cache/compare/v4...v5) --- updated-dependencies: - dependency-name: actions/upload-artifact dependency-version: '6' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions - dependency-name: actions/cache dependency-version: '5' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions ... Signed-off-by: dependabot[bot] <[email protected]> * Fix premature assignment expiration warning (#6489) * CI: enable sccache (#6473) This avoids doing redundant work by caching compiler results. Because CI does less work, we can run more redundant CI jobs and avoid the mental overhead of manually de-duping jobs. * fix: resurrect LoadShareDesignDialog.h which got lost during merge from OpenSCAD * fix: return type of findModelObject which got inadvertently changed during merge from OpenSCAD * fix: version handling which got broken during merge from OpenSCAD * fix: add missing conv. for editor file path in parseDocument which got lost during OpenSCAD merge * fix: correct variable names in argument parsing for python_cylinder function after OpenSCAD merge * fix: revert test result for use-tests-expected which was inadvertently changed during OpenSCAD merge * fix: update argument parsing for python_cylinder function after OpenSCAD merge * fix: correct discretizer handling in python_cylinder after OpenSCAD merge The OpenSCAD merge introduced CurveDiscretizer to centralize fn/fa/fs parameter handling. However, python_cylinder was creating the discretizer twice: once before argument parsing (which removes fn/fa/fs from kwargs) and again when creating the CylinderNode (from already-modified kwargs). This caused cylinders to ignore fn/fa/fs parameters and render with incorrect facet counts (e.g., 6 sides instead of specified 12). Fixed by: - Removing fn/fa/fs from kwlist (handled by CreateCurveDiscretizer) - Removing unused fn/fa/fs local variables - Using single discretizer instance throughout - Updating format string to match reduced parameter list Resolves incorrect cylinder rendering in hinge and spline tests. Co-Authored-By: Claude Sonnet 4.5 <[email protected]> * fix: add merge-mode-functions flag to gcovr to fix coverage collection After the OpenSCAD merge, gcovr encounters functions in AST.h that appear on multiple lines across different compilation units, causing coverage merging to fail with: GcovrMergeAssertionError: Got function <unknown function> on multiple lines Adding --merge-mode-functions=merge-use-line-0 tells gcovr to use line 0 as the canonical line for functions when merging, resolving the conflict. This fixes the Linux CI coverage step failures while maintaining the same test results (all tests still pass). Co-Authored-By: Claude Sonnet 4.5 <[email protected]> --------- Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: Nicolas Vaagen <[email protected]> Co-authored-by: Torsten Paul <[email protected]> Co-authored-by: Jordan Brown <[email protected]> Co-authored-by: Marius Kintel <[email protected]> Co-authored-by: Ma-XX-oN <[email protected]> Co-authored-by: AaronVerDow <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: coryrc <[email protected]> Co-authored-by: Alex Kent Hajnal <[email protected]> Co-authored-by: Sohler Günther <[email protected]> Co-authored-by: Nicolas Vaagen <[email protected]> Co-authored-by: Glebs Ivanovskis <[email protected]> Co-authored-by: pakhare <[email protected]> Co-authored-by: Flaviu Tamas <[email protected]> Co-authored-by: Guenther Sohler <[email protected]> Co-authored-by: Claude Sonnet 4.5 <[email protected]> * docs: add CLAUDE.md for project guidance and development workf (#288) * feat(python): make add_parameter return value and add feature toggle (#290) - Modified add_parameter() to return the parameter value instead of None - Added feature toggle 'add-parameter-pure-function' to control global variable creation - When feature is disabled (default): returns value AND creates global variable (backward compatible) - When feature is enabled: returns value WITHOUT creating global variable (pure function) - Added two test cases to verify both behaviors This allows gradual deprecation of global variable creation while maintaining backward compatibility. Co-authored-by: Claude Sonnet 4.5 <[email protected]> * chore(master): release 0.8.8 (#289) * fix: evalauting fn for path_extrude again (#287) * fix: evaluating fn for path_extrude again * fix: also wrap works again --------- Co-authored-by: Guenther Sohler <[email protected]> * feat: add children() function * fix(rpm): resolve RPM build failures by forcing static linking (#341) * fix(rpm): force static linking to match Debian package behavior RPM builds were creating shared libraries (libOpenSCADPy.so, libmanifold.so.3, libClipper2.so.1) but not installing them, causing runtime dependency failures. Debian builds statically link these libraries into the main binary, resulting in a working package. Root cause: Fedora's %cmake macro historically sets BUILD_SHARED_LIBS=ON by default, overriding the project's BUILD_SHARED_LIBS=OFF setting. Debian builds don't use this macro and respect the project default. Changes: - Add -DBUILD_SHARED_LIBS=OFF to CMake flags to force static linking - Remove libClipper2.so, libOpenSCADPy.so, libmanifold.so from %files section (these are now statically linked) - Update __requires_exclude to only filter libfive.so (the only bundled shared library that is actually installed) This matches Debian's proven working approach and produces a self-contained binary with only libfive.so as a separate library. Fixes build failures in v0.8.20-v0.8.23 and broken v0.8.19 packages. Co-Authored-By: Claude Sonnet 4.5 <[email protected]> * fix(cmake): use CMAKE_INSTALL_LIBDIR for libfive install path The libfive install rule was using hardcoded "lib" destination, but on 64-bit systems libraries should go to lib64. This caused RPM packaging to fail because the spec file expects libraries in %{_libdir} (which expands to lib64). Changed from: install(FILES ... DESTINATION "lib") To: install(FILES ... DESTINATION ${CMAKE_INSTALL_LIBDIR}) This matches how Debian's rules use CMAKE_INSTALL_RPATH=/usr/lib and ensures libfive.so is installed to the correct architecture- specific library directory. Co-Authored-By: Claude Sonnet 4.5 <[email protected]> --------- Co-authored-by: Claude Sonnet 4.5 <[email protected]> * feat: add installation documentation to website (#342) * feat: add installation documentation to website * feat(web): add installation instructions for GNU Guix * chore(master): release 0.8.24 (#339) * fix(ci): resolve RPM signing passphrase issue (#345) The RPM signing step was failing because the GPG passphrase wasn't being correctly passed through stdin to the rpm/gpg subprocess. The previous approach using --passphrase-fd 0 didn't work in the GitHub Actions container environment. This fix: - Creates a temporary file for the GPG passphrase instead of piping - Configures the %__gpg_sign_cmd RPM macro to use --passphrase-file - Uses rpmsign instead of rpm for clarity - Properly cleans up the passphrase file after signing The passphrase file approach is more reliable across different environments and avoids the stdin redirection issues. Fixes: https://github.com/pythonscad/pythonscad/actions/runs/20981164976 * fix(ci): AppImage libfive dependency, Qt5 support, ARM64 support, and distribution updates (#347) * fix(ci): Fix EOL date for Debian Trixie * ci(debian): stop building packages for Ubuntu 24.10 oracular as it's EOL * ci(debian): fix EOL date for Ubuntu 22.04 LTS jammy * ci(debian): fix EOL date for Ubuntu 24.04 LTS noble * fix(ci): resolve AppImage libfive dependency and add Qt5/ARM64 support This commit fixes the AppImage build and adds multi-architecture support: 1. **Libfive dependency issue**: The AppImage build started failing after commit 50591251c which changed libfive install path from hardcoded "lib" to CMAKE_INSTALL_LIBDIR. Root cause: - On Ubuntu 24.04, CMAKE_INSTALL_LIBDIR defaults to "lib/x86_64-linux-gnu" - libfive.so is now installed to usr/lib/x86_64-linux-gnu/libfive.so - linuxdeploy expects libraries in usr/lib/ and can't find it Error: ``` ERROR: Could not find dependency: libfive.so ERROR: Failed to deploy dependencies for existing files ``` Solution: - Override CMAKE_INSTALL_LIBDIR=lib for AppImage builds - This keeps libfive.so in usr/lib/ where linuxdeploy expects it - Appropriate for AppImage which uses a simplified /usr layout 2. **Add Qt5 support**: Converted the AppImage workflow to a matrix build that creates both Qt5 and Qt6 versions of the AppImage. 3. **Add ARM64 (aarch64) support**: Extended the matrix to build for both x86_64 and aarch64 architectures. Implementation: - Added architecture detection (auto-detect via uname -m or ARCH env var) - Download correct architecture versions of linuxdeploy, appimagetool, and linuxdeploy-plugin-qt - Use QEMU emulation for ARM64 builds on x86_64 runners - Include architecture and Qt version in AppImage filename to avoid naming conflicts Changes: - Added ARCH environment variable to create_appimage.sh - Script auto-detects architecture or uses ARCH environment variable - Workflow builds 4 variants: Qt5/Qt6 × x86_64/aarch64 - Each variant uses the appropriate dependency profile - AppImage filenames include Qt version and architecture: - PythonSCAD-<version>-qt5-x86_64.AppImage - PythonSCAD-<version>-qt5-aarch64.AppImage - PythonSCAD-<version>-qt6-x86_64.AppImage - PythonSCAD-<version>-qt6-aarch64.AppImage Fixes: https://github.com/pythonscad/pythonscad/actions/runs/20981164946 * feat: add ARM64 AppImage support using Docker containers Implements ARM64 (aarch64) AppImage builds using Docker containers, following the same pattern as Debian package builds. This approach uses multi-arch Docker images with QEMU emulation at the container level, allowing proper execution of native ARM64 tools. Changes: - Add arch matrix dimension (x86_64, aarch64) to workflow - Use ubuntu:24.04 Docker container for ARM64 builds - Keep native builds for x86_64 (no container overhead) - Split dependency installation for native vs container contexts - Update artifact naming to include architecture - Pass ARCH environment variable to build script This enables building AppImages for both x86_64 and ARM64 across Qt5 and Qt6, resulting in 4 total AppImage variants. Co-Authored-By: Claude Sonnet 4.5 <[email protected]> * fix(ci): use explicit docker run for ARM64 AppImage builds The previous approach using the container field didn't work because GitHub Actions doesn't support specifying platform/architecture for the container image. This caused Docker to pull the amd64 image even for ARM64 builds, resulting in "cannot execute binary file" errors. Solution: - x86_64 builds: Continue running natively on ubuntu-24.04 runner - aarch64 builds: Use explicit 'docker run --platform linux/arm64' command after setting up QEMU This approach ensures the ARM64 Docker container pulls the correct arm64 image and all tools inside run as native ARM64 binaries. Co-Authored-By: Claude Sonnet 4.5 <[email protected]> * fix(ci): fix ARM64 builds for Debian and RPM packages Both Debian and RPM workflows had the same issue as the AppImage workflow: using the container field pulls amd64 images even for ARM64 matrix entries, resulting in mislabeled packages (amd64 packages with arm64 in filename). Issue discovered: - Debian packages named *_arm64.deb were actually Architecture: amd64 - This occurred because container field doesn't respect platform/arch - QEMU setup never ran due to condition: matrix.qemu && !matrix.docker_image Solution (same approach as AppImage): - amd64 builds: Continue using container field (existing behavior) - arm64 builds: Use explicit 'docker run --platform linux/arm64' - Set up QEMU before ARM64 builds - Wrap all build steps (dependencies, build, signing) in docker run This ensures ARM64 packages are actually built for ARM64 architecture. Testing: - Previous v0.8.24 arm64.deb packages are actually amd64 - New builds will produce genuine ARM64 binaries Co-Authored-By: Claude Sonnet 4.5 <[email protected]> * fix(ci): escape multi-line secrets in ARM64 RPM build script The ARM64 RPM build was failing with 'unexpected EOF while looking for matching quote' error. This was caused by multi-line secret values (GPG keys and passphrases) being substituted into echo commands with double quotes inside a bash -c '...' single-quoted string. When GitHub Actions substitutes a multi-line secret like DEB_SIGNING_KEY into: echo "${{ secrets.DEB_SIGNING_KEY }}" inside bash -c '...', it creates: echo "-----BEGIN PGP... line2 line3..." which breaks bash syntax with unclosed quotes. Solution: Use single quotes around the secrets (with proper escaping for the bash -c context) to preserve them literally: echo '\''...'\'\' This matches the pattern already used successfully in the Debian workflow. Fixes: https://github.com/pythonscad/pythonscad/actions/runs/20997281343 * fix(ci): add docker images for Ubuntu jammy/noble ARM64 builds Ubuntu jammy and noble had docker_image set to null because they use native GitHub runners for amd64 builds. However, ARM64 builds require Docker with the --platform flag, which needs a valid image name. When docker_image was null, the Docker command was malformed: docker run ... ${{ matrix.docker_image }} bash -c ... docker run ... bash -c ... # docker_image=null becomes empty string Docker then interpreted 'bash' as the image name and pulled bash:latest, which doesn't have apt-get, causing the build to fail. Fix: Add explicit docker_image values (ubuntu:jammy, ubuntu:noble) so ARM64 builds work correctly. The amd64 builds still use native runners via the runner field. Fixes: https://github.com/pythonscad/pythonscad/actions/runs/20997278400 * ci: do not build for arm64 at the moment as it is too slow * chore: remove ARM64 from repository website templates ARM64 builds are temporarily disabled due to slow QEMU emulation. Updated HTML templates in update-apt-repo.sh and update-yum-repo.sh to remove arm64/aarch64 from supported architectures list. --------- Co-authored-by: Claude Sonnet 4.5 <[email protected]> * chore(master): release 0.8.25 (#346) * feat(web): improve releases page and add cheat sheet (#348) * feat(ci): implement draft release workflow with auto-publish - Add 'draft: true' to release-please config to create draft releases - Create publish-release workflow that monitors build workflows - Auto-publish when all builds succeed - Keep draft and add status report when builds fail - Ensures downloads page only shows complete releases * fix: fix cheatsheet URL to adapt to new website layout * fix(web): correct HTML structure in cheatsheet - Fix missing closing </div> tags in multiple sections - Change incorrect </section> tags to </div> - Fix <<div typo to <div> - Add missing <div class='grid'> wrappers - Add missing closing </code></pre> tags - Fixes layout issues where sections overlapped or displayed incorrectly * docs: add cheat-sheet page to website * chore(master): release 0.8.26 (#349) * fix(ci): use prerelease instead of draft for release workflow (#351) Problem: Draft releases don't trigger the 'release: [created]' event, so build workflows weren't running after release-please created releases. Solution: Use 'prerelease: true' instead of 'draft: true': - Prereleases DO trigger the release:created event - Build workflows now trigger properly - Publish workflow clears prerelease flag when all builds succeed - Users still won't see incomplete releases (prereleases are hidden) Changes: - .release-please-config.json: Changed draft:true to prerelease:true - .github/workflows/publish-release.yml: Updated to handle prereleases - Check for prerelease flag instead of draft - Clear prerelease flag when publishing - Updated error messages * chore: update from OpenSCAD (#344) * Run once daily. * Update (and run) translation scripts; require bash; enable error options. * Bumped Clipper to 2.0.1 (#6494) * Bumped Clipper to 2.0.1 * Add bugfix for https://github.com/AngusJohnson/Clipper2/pull/1054 --------- Co-authored-by: Torsten Paul <[email protected]> Co-authored-by: Marius Kintel <[email protected]> * chore: remove obsolete Python 3.11 binaries (#353) * chore: remove uncrustify config as it is no longer in use (#355) * chore(master): release 0.8.27 (#356) * fix(ci): fix GPG signing and unify build matrices (#357) * fix(ci): fix GPG signing in non-TTY environments The RPM signing was failing in GitHub Actions with: - "Could not set GPG_TTY to stdin: Inappropriate ioctl for device" - "gpg: signing failed: Bad passphrase" Root causes identified and fixed: 1. Duplicate gpg in macro: `%{__gpg} gpg` expanded to `/usr/bin/gpg gpg` - Fixed by removing the duplicate `gpg` 2. Wrong option order: When `--pinentry-mode loopback` came first, GPG didn't recognize subsequent options like `--batch` - Fixed by putting `--batch` first in the options list 3. Missing gpg-agent configuration: The agent needs `allow-loopback-pinentry` in gpg-agent.conf to allow passphrase injection without a TTY - Fixed by creating gpg-agent.conf before importing keys 4. Using long-form options: Changed `--sign --detach-sign --output` to `-sbo` (the short form used by rpm's default macro) 5. Set GPG_TTY="" to suppress harmless TTY warnings in CI Tested in Docker without TTY to confirm the fix works before applying to the workflow. Co-Authored-By: Claude Opus 4.5 <[email protected]> * refactor(ci): read RPM build matrix from supported-distributions.json - Add Fedora 42, Fedora 43, and Rocky Linux 9 (EL) entries to supported-distributions.json - Refactor build-rpm-packages.yml to use prepare-matrix job that reads from the JSON file, matching the pattern used by build-debian-packages.yml - Both workflows now share a single source of truth for supported distributions and architectures - Update notes section to document supported families Co-Authored-By: Claude Opus 4.5 <[email protected]> * chore: remove arm64 architecture support for Fedora 42 and 43 Drop arm64 from the supported architectures list for Fedora 42 and 43 distributions as build times are currently way too long. --------- Co-authored-by: Claude Opus 4.5 <[email protected]> * fix(web): use theme CSS variables for cheatsheet dark mode support (#358) Replace hardcoded light-mode colors with Material for MkDocs CSS variables so the cheatsheet page correctly adapts to dark mode. Co-authored-by: Claude Opus 4.5 <[email protected]> * chore(master): release 0.8.28 (#359) * fix(ci): import GPG public key to RPM keyring for signature verification (#361) * chore(master): release 0.8.29 (#362) * fix(ci): fix RPM signing for Fedora 43 (RPM 6.x) (#363) RPM 6.0 introduced breaking changes where custom %__gpg_sign_cmd macro overrides no longer work due to parametric macros. This commit adds version-aware signing: - RPM 4.x (Fedora 42, Rocky 9): Use existing approach with %__gpg_sign_cmd - RPM 6.x (Fedora 43): Use gpg-preset-passphrase to cache passphrase in gpg-agent and use %_openpgp_sign_id macro Fixes signing failure on Fedora 43 while maintaining compatibility with older RPM versions. * chore(master): release 0.8.30 (#364) * Update lodepng to 20260102 and add a README * Add to readme where the lexertl source came from Wrote a bash script to confirm this, but it was HEAD of the other repo at the time of the original commit. * Estimate cache size for manifold objects (#6499) * clang-format: Add AllowShortFunctionsOnASingleLine: InlineOnly (#6530) * Correctly preferences after OpenSCADApp has finished initializing (#6533) * Add hidapi README (#6532) Based on comments to https://github.com/openscad/openscad/issues/6513 add a README.md for hidapi like the other external libraries. * Add color list window. * Add support for double-click to insert the color name / color() call. * Ensure newline after HtmlLink message; Skip messages in ErrorList window. * Require "xkcd:" prefix to switch to the XKCD color map. * Suppress log messages of type HtmlLink on command line. * Handle API change between Qt5 and Qt6 (qreal -> float) in QColor::getHslF(). * fix: used other version from regression * fix:udpated tests * fix(ci): fix keygrip extraction for RPM 6.x signing (#366) The previous grep pattern '^\s*sec' expected whitespace before "sec" but GPG output starts with just "sec" (no leading whitespace). This caused the keygrip extraction to fail on Fedora 43 (RPM 6.x). Fixed by using awk to find lines starting with "sec" and then extracting the keygrip from the following line containing "Keygrip". Also added debug output to show the full GPG key details for easier troubleshooting. Co-authored-by: Claude Opus 4.5 <[email protected]> * chore(master): release 0.8.31 (#367) * chore: merge 91 commits from openscad/master (#277) * add correct StartupWMClass to set the correct icon on Gnome I believe this is the XDG way to do this, but I am not sure. * Check for save/discard/cancel on quit. Fixes #3971. (#6311) * Bump CGAL to 6.1.0 (#6318) * Make vector swizzle experimental (#6307) * Don't log to stderr (#6323) * Fully separate zoom levels for each tab. Fixes #4742. (#6320) * Fix/cleanup compiler warnings openscad/openscad#6245 (#6300) * Remove obsolete HTML example converter (#6328) * reformat python tests scripts in pep8 (#6330) * Bump the github-actions group with 3 updates Bumps the github-actions group with 3 updates: [actions/upload-artifact](https://github.com/actions/upload-artifact), [github/codeql-action](https://github.com/github/codeql-action) and [actions/stale](https://github.com/actions/stale). Updates `actions/upload-artifact` from 4 to 5 - [Release notes](https://github.com/actions/upload-artifact/releases) - [Commits](https://github.com/actions/upload-artifact/compare/v4...v5) Updates `github/codeql-action` from 3 to 4 - [Release notes](https://github.com/github/codeql-action/releases) - [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md) - [Commits](https://github.com/github/codeql-action/compare/v3...v4) Updates `actions/stale` from 9 to 10 - [Release notes](https://github.com/actions/stale/releases) - [Changelog](https://github.com/actions/stale/blob/main/CHANGELOG.md) - [Commits](https://github.com/actions/stale/compare/v9...v10) --- updated-dependencies: - dependency-name: actions/upload-artifact dependency-version: '5' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions - dependency-name: github/codeql-action dependency-version: '4' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions - dependency-name: actions/stale dependency-version: '10' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions ... Signed-off-by: dependabot[bot] <[email protected]> * Check print service name only if remote print service is selected. This fixes the issue that the print dialog is always shown if the local application is selected for printing. * Don't crash MouseSelector, debug GL init (#6334) * beautify (#6342) * Set QSCINTILLA_DLL for MINGW (and not MXECROSS). (#6343) * Warn on offset() with both r and delta specified (#6348) * Fix QFile::Open nodiscard warning with Qt6 6.10.0. (#6364) * Update cmdline help to show Manifold as default (#6350) * Rm unused function. Fixes #6352. (#6353) * Add Color module spec (#6339) * Enables "Ctrl/Cmd-Mouse-wheel zooms text" config option to work (Preliminary fix for #6368) * Entirely removed superfluous EditorInterface::wheelEvent method * Bump Qt to 6.5.7 (#6371) * CircleCI: Update macOS resources to 26.1 on M4 (#6324) * Refactor Discretization Special Usage (#6251) * Tessallation specials $fn$fs$fa now grouped in an object Instead of fn,fs,fa as three doubles, they are passed as an object packaged up with the methods on them. * Repair basic functionality (#6376) * Fix indentation of CC0 template text. (#6375) Also clean up line endings. Add templates/README.md * Fix a number of non-ASCII file name issues (#6136) * Try caching MSYS2 packages Fixes #6381 if it works * Fix `if` in yaml wrong * Don't want `if`s anyway doesn't matter I did them wrong * Update msys2 gh actions caching Make one new cache a day from scratch. Leave note why we don't find the most recent cache. Skip saving since caches are immutable. * Comment updates + build up cache over a month * We can read the determined msys2 directory We're not restoring before the setup, like I originally planned. * Updating script to output list of pacboy packages to workflow Easier than my other approach, actually. Ah, whatever. * Don't change $MSYSTEM in GH Actions * Oops, I need to specify the bash shell I think * Use arrays instead of backslash line termination * Don't fail fast because it fails due to caching a lot well, until we actually get one * std::wstring_convert and std::codecvt_utf8_utf16 is going to be depreciated. Use boost::nowide::narrow (#6356) * Add Unit tests to fix measuring (#6299) * Starting point of release build CI workflow (#6374) * Define executors separately * Added OpenSCAD release build job with create-tarball * Centralize version string management (#6388) * Centralize version string management * Identify build tag in OPENSCAD_VERSION, and correctly extract OPENSCAD_SHORTVERSION * cimg/python: no 'latest' tag. Use 3.14 (#6399) * Disable WIP macOS release build (#6398) * CircleCI Bugfix: Use OPENSCAD_VERSION instead of VERSION (#6400) * create-tarball: Amend files manually since git-archive-all doesn't consider unversioned files (#6401) * cd to /tmp/artifacts before creating shasums (#6402) * Add CMake support + shell helpers for VERSION.txt and COMMIT.txt (#6403) * Ingest VERSION.txt and COMMIT.txt at cmake time * Shell helper for establishing version from VERSION.txt and COMMIT.txt * Snapshot builds now uses the version helpers to unify version management, also deprecated 32-bit windows (#6404) * Bugfix: binary name on macOS needs correct case (#6409) * Parametrize macOS build for release vs. snapshot (#6406) * Parametrize macOS build for release vs. snapshot * Use establish_version.sh and don't require git * Always checkout in the same way; with submodules * Bump actions/checkout from 5 to 6 in the github-actions group (#6418) Bumps the github-actions group with 1 update: [actions/checkout](https://github.com/actions/checkout). Updates `actions/checkout` from 5 to 6 - [Release notes](https://github.com/actions/checkout/releases) - [Changelog](https://github.com/actions/checkout/blob/main/CHANGELOG.md) - [Commits](https://github.com/actions/checkout/compare/v5...v6) --- updated-dependencies: - dependency-name: actions/checkout dependency-version: '6' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * Delete discretizer keyword args PyArg_ParseTupleAndKeywords doesn't allow keyword args it doesn't know about, so we delete them if found. Manually testing the fix. * Fix formatting of "Browser" header (#6433) Markdown rendered is confused by non-breaking space (`U+A0`) - replaced with a regular space. * Make find highlighting case-insensitive * chore: codegen version into manpage (#6448) * chore: codegen version into manpage testing process: - created a VERSION.txt file - ran the CMake build - confirm that the output `build/doc/openscad.1` has the same version as VERSION.txt * remove extra comment * Added job parameters openscad_build_type and openscad_platform (#6449) * Remove deprecated child() and assign(). (#6428) * Remove deprecated child() and assign(). Fixes #5557. Fixes #6425. * Update releases/next.md: child(), assign(), object(), textmetrics(), fontmetrics Remove deprecated child() and assign(). Add object(). Reclassify textmetrics() and fontmetrics() as functions. * More CircleCI cleanup (#6450) * Bump manifold to 3.3.2 and onetbb to 2022.3.0 (#6455) * Added failing test for #6456 (#6459) * Bumped manifold to fix #6235 (#6467) * Temporarily restore exact job names used by circleci-download-artifacts.py (#6469) * CircleCI: Update name of wasm build (#6470) * CircleCI release builds (#6412) * Move version info to .c file (#6463) * chore: bump nix-shell for latest version * convert to flake (for lockfile) * use stable release * remove duplicate lock entry * Add warning to cmake output that ENABLE_CGAL=OFF is a dev-only flag. * Fix font sample rendering in font list window. * Remove obsolete FontListDialog. * Change console font size via mouse wheel event (fixes #6186). This is only changing the current console widget, not the value in the preferences. * Align naming of release jobs, correct naming of tarball (#6484) * Bugfix: Forgot to update dependencies on tarball (#6485) * Bump the github-actions group with 2 updates Bumps the github-actions group with 2 updates: [actions/upload-artifact](https://github.com/actions/upload-artifact) and [actions/cache](https://github.com/actions/cache). Updates `actions/upload-artifact` from 5 to 6 - [Release notes](https://github.com/actions/upload-artifact/releases) - [Commits](https://github.com/actions/upload-artifact/compare/v5...v6) Updates `actions/cache` from 4 to 5 - [Release notes](https://github.com/actions/cache/releases) - [Changelog](https://github.com/actions/cache/blob/main/RELEASES.md) - [Commits](https://github.com/actions/cache/compare/v4...v5) --- updated-dependencies: - dependency-name: actions/upload-artifact dependency-version: '6' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions - dependency-name: actions/cache dependency-version: '5' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions ... Signed-off-by: dependabot[bot] <[email protected]> * Fix premature assignment expiration warning (#6489) * CI: enable sccache (#6473) This avoids doing redundant work by caching compiler results. Because CI does less work, we can run more redundant CI jobs and avoid the mental overhead of manually de-duping jobs. * fix: resurrect LoadShareDesignDialog.h which got lost during merge from OpenSCAD * fix: return type of findModelObject which got inadvertently changed during merge from OpenSCAD * fix: version handling which got broken during merge from OpenSCAD * fix: add missing conv. for editor file path in parseDocument which got lost during OpenSCAD merge * fix: correct variable names in argument parsing for python_cylinder function after OpenSCAD merge * fix: revert test result for use-tests-expected which was inadvertently changed during OpenSCAD merge * fix: update argument parsing for python_cylinder function after OpenSCAD merge * fix: correct discretizer handling in python_cylinder after OpenSCAD merge The OpenSCAD merge introduced CurveDiscretizer to centralize fn/fa/fs parameter handling. However, python_cylinder was creating the discretizer twice: once before argument parsing (which removes fn/fa/fs from kwargs) and again when creating the CylinderNode (from already-modified kwargs). This caused cylinders to ignore fn/fa/fs parameters and render with incorrect facet counts (e.g., 6 sides instead of specified 12). Fixed by: - Removing fn/fa/fs from kwlist (handled by CreateCurveDiscretizer) - Removing unused fn/fa/fs local variables - Using single discretizer instance throughout - Updating format string to match reduced parameter list Resolves incorrect cylinder rendering in hinge and spline tests. Co-Authored-By: Claude Sonnet 4.5 <[email protected]> * fix: add merge-mode-functions flag to gcovr to fix coverage collection After the OpenSCAD merge, gcovr encounters functions in AST.h that appear on multiple lines across different compilation units, causing coverage merging to fail with: GcovrMergeAssertionError: Got function <unknown function> on multiple lines Adding --merge-mode-functions=merge-use-line-0 tells gcovr to use line 0 as the canonical line for functions when merging, resolving the conflict. This fixes the Linux CI coverage step failures while maintaining the same test results (all tests still pass). Co-Authored-By: Claude Sonnet 4.5 <[email protected]> --------- Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: Nicolas Vaagen <[email protected]> Co-authored-by: Torsten Paul <[email protected]> Co-authored-by: Jordan Brown <[email protected]> Co-authored-by: Marius Kintel <[email protected]> Co-authored-by: Ma-XX-oN <[email protected]> Co-authored-by: AaronVerDow <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: coryrc <[email protected]> Co-authored-by: Alex Kent Hajnal <[email protected]> Co-authored-by: Sohler Günther <[email protected]> Co-authored-by: Nicolas Vaagen <[email protected]> Co-authored-by: Glebs Ivanovskis <[email protected]> Co-authored-by: pakhare <[email protected]> Co-authored-by: Flaviu Tamas <[email protected]> Co-authored-by: Guenther Sohler <[email protected]> Co-authored-by: Claude Sonnet 4.5 <[email protected]> * feat: multmatrix operates on Vector3d * fix:removed wrong printf * added lasercutr lib * support #2 cuts * feat: start allow to alter existing solid paramters * feat(ci): add native Windows packaging workflow (#371) * feat(ci): add native Windows packaging workflow Add a new GitHub Actions workflow that builds PythonSCAD natively on Windows using MSYS2, packages it with CPack (ZIP + NSIS installer), and runs a sanity test (pythonscad --info) to verify the executable works before release. This addresses issue #360 where MXE cross-compiled builds are missing MinGW runtime DLLs (libgcc_s_seh-1.dll), causing the application to fail to launch on Windows. The native MSYS2 build correctly bundles all required DLLs. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(ci): use workspace-relative path for artifacts Changed from /tmp/out/ (MSYS2-only path) to artifacts/ (workspace-relative) to fix upload-artifact action which runs in native Windows context. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(ci): bundle MSYS2 runtime DLLs in package Add step to use ldd to find and copy all MSYS2 DLL dependencies recursively into the staging directory before packaging. This fixes missing DLL errors: - libgcc_s_seh-1.dll - libcairo-2.dll - libwinpthread-1.dll - libstdc++-6.dll And any other transitive dependencies. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(ci): rewrite DLL bundling to avoid bash subshell issues Changed from recursive function to iterative loop to avoid bash 'local' keyword failures in subshells created by pipes. Uses process substitution instead of pipes and iterates up to 10 times to find all transitive DLL dependencies. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(ci): use CMake's install(RUNTIME_DEPENDENCY_SET) for DLL bundling Instead of manually bundling DLLs with ldd, use CMake 3.21+'s native runtime dependency resolution. This ensures CPack packages include all required MSYS2 runtime DLLs (libgcc_s_seh-1.dll, libstdc++-6.dll, etc.). The RUNTIME_DEPENDENCY_SET feature automatically: - Finds all DLL dependencies of the executable - Excludes Windows system DLLs (api-ms-*, system32, etc.) - Installs them to the package destination This is the proper CMake-based solution rather than shell scripting. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(ci): add Qt plugins for native MSYS2 Windows builds Qt applications require platform plugins (qwindows.dll) to start. Also install style, imageformat, and iconengine plugins for full functionality. Plugins installed: - platforms/qwindows.dll (mandatory) - styles/*.dll (native Windows look) - imageformats/*.dll (image loading) - iconengines/*.dll (SVG icons) Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(ci): add qt.conf for Qt plugin discovery on Windows Qt needs a qt.conf file to know where to find plugins when they're not in the default 'plugins/' subdirectory. This file tells Qt that the plugin directories (platforms/, imageformats/, etc.) are in the same directory as the executable. Contents of qt.conf: [Paths] Plugins = . This should fix the missing toolbar icons issue. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(windows): bundle Qt SVG runtime DLL * fix(windows): use MSYS2 Python stdlib instead of embed package For native MSYS2 Windows builds, the executable is linked against MSYS2's libpython3.13.dll. The RUNTIME_DEPENDENCY_SET copies this DLL to the install directory, but the Python standard library was being installed from python.org's embed package, which has an incompatible structure (python313.dll + python313.zip vs libpython3.13.dll + lib/python3.13/). This mismatch caused Python initialization to fail with: "Could not find platform independent libraries <prefix>" "ERROR: Python 3.13.11 not found. Is it installed?" Fix by installing the MSYS2 Python standard library from the system installation, which matches the ABI of the linked libpython DLL. Co-Authored-By: Claude Opus 4.5 <[email protected]> * ci: disable MXE cross compiled Windows builds They are currently broken and are being replaced by native MSYS2 builds. * feat(windows): improve NSIS installer with start menu, launch option, and multiuser support - Fix NSIS configuration to work for both MXE cross-compile and native MSYS2 builds - Add explicit start menu shortcut creation in PythonSCAD folder and root - Register .scad file association with pythonscad.exe instead of openscad.exe - Add "Run PythonSCAD" checkbox on installer finish page - Add multiuser.nsh for per-user installation support infrastructure - Update URLs and icon references from OpenSCAD to PythonSCAD - Properly clean up shortcuts and file associations on uninstall Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(windows): convert NSIS include paths to native Windows format NSIS requires backslashes for include paths on Windows. Using file(TO_NATIVE_PATH) to convert CMake paths to native format. Also removes unused multiuser.nsh include that had unmet dependencies. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(windows): use relative paths for NSIS includes CPack runs NSIS from build/_CPack_Packages/<system>/NSIS/, so use relative paths (../../../) to reference files in the build root. Also copies mingw-file-association.nsh to build dir for consistency. Co-Authored-By: Claude Opus 4.5 <[email protected]> * ci(windows): remove experimental suffix from native build artifacts The MSYS2-based native Windows build is now the primary Windows build, so artifact filenames no longer need the msys2-experimental suffix. Co-Authored-By: Claude Opus 4.5 <[email protected]> * chore: disable pre-major version bump settings in release-please Configure release-please to use standard semver bumping by setting bump-minor-pre-major and bump-patch-for-minor-pre-major to false --------- Co-authored-by: Claude Opus 4.5 <[email protected]> * feat(python): extend add_parameter() with customizer options (#368) * fix: memory bugs in add_parameter Python string handling PyUnicode_AsEncodedString creates temporary bytes objects. PyBytes_AS_STRING returns pointer to internal buffer. Without proper cleanup, this creates dangling pointers causing segfaults. Fixed by copying to std::string before calling Py_DECREF(). Affects: default string values, options lists, and dict labels. * fix: check event type before casting to QKeyEvent Static cast of QEvent to QKeyEvent without checking event type causes heap-use-after-free when event is not actually a key event. Fixed by checking event->type() == QEvent::KeyPress before cast. * fix: heap-use-after-free in add_parameter annotations Assignment::addAnnotations stores raw pointers to Annotation objects. Stack-allocated AnnotationList was being destroyed when function returned, leaving dangling pointers. Fixed by heap-allocating AnnotationList with new, matching the pattern used in CommentParser.cc. * fix(gui): fix memory bugs in customizer parameter handling - ParameterWidget: Fix use-after-free when slider is dragged during F5 re-render - Use deleteLater() for widget cleanup to handle pending Qt events - Keep old parameters alive until widgets are fully deleted - Process events after scheduling deletion to handle in-flight events - MainWindow: Fix crash on window close due to RubberBandManager access - Add isBeingDestroyed flag to guard eventFilter during destruction - Set flag at start of destructor before any cleanup * test: add expected output for add_parameter extended tests * fix(ci): import GPG public key to RPM keyring for signature verification (#361) * Add color list window. * Add support for double-click to insert the color name / color() call. * Suppress log messages of type HtmlLink on command line. * fix: used other version from regression * fix:udpated tests * path points is not writable * sync to master * fix; sync regression * sycn with master * fix: convert face to polyghons when not already * fix: improve lasercut library * fix: face/face cuts working now * fix: create also lasercutmodels from whole shapes instead from faces * fix: returning the final result rather than showing it * make sure, that face2face joints get better slices * feat: implemented export GCODE also pylaser.py got a maximum result size parameter * fix: add missed file --------- Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: Guenther Sohler <[email protected]> Co-authored-by: nomike <[email protected]> Co-authored-by: Nicolas Vaagen <[email protected]> Co-authored-by: Torsten Paul <[email protected]> Co-authored-by: Jordan Brown <[email protected]> Co-authored-by: Marius Kintel <[email protected]> Co-authored-by: Ma-XX-oN <[email protected]> Co-authored-by: AaronVerDow <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: coryrc <[email protected]> Co-authored-by: Alex Kent Hajnal <[email protected]> Co-authored-by: Nicolas Vaagen <[email protected]> Co-authored-by: Glebs Ivanovskis <[email protected]> Co-authored-by: pakhare <[email protected]> Co-authored-by: Flaviu Tamas <[email protected]> Co-authored-by: Claude Sonnet 4.5 <[email protected]> Co-authored-by: pythonscad-gh-bot <[email protected]>

* feat: more properties in __getitem__ * chore: cleanup * chore: add tests * chore: merge 91 commits from openscad/master (#277) * add correct StartupWMClass to set the correct icon on Gnome I believe this is the XDG way to do this, but I am not sure. * Check for save/discard/cancel on quit. Fixes #3971. (#6311) * Bump CGAL to 6.1.0 (#6318) * Make vector swizzle experimental (#6307) * Don't log to stderr (#6323) * Fully separate zoom levels for each tab. Fixes #4742. (#6320) * Fix/cleanup compiler warnings openscad/openscad#6245 (#6300) * Remove obsolete HTML example converter (#6328) * reformat python tests scripts in pep8 (#6330) * Bump the github-actions group with 3 updates Bumps the github-actions group with 3 updates: [actions/upload-artifact](https://github.com/actions/upload-artifact), [github/codeql-action](https://github.com/github/codeql-action) and [actions/stale](https://github.com/actions/stale). Updates `actions/upload-artifact` from 4 to 5 - [Release notes](https://github.com/actions/upload-artifact/releases) - [Commits](https://github.com/actions/upload-artifact/compare/v4...v5) Updates `github/codeql-action` from 3 to 4 - [Release notes](https://github.com/github/codeql-action/releases) - [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md) - [Commits](https://github.com/github/codeql-action/compare/v3...v4) Updates `actions/stale` from 9 to 10 - [Release notes](https://github.com/actions/stale/releases) - [Changelog](https://github.com/actions/stale/blob/main/CHANGELOG.md) - [Commits](https://github.com/actions/stale/compare/v9...v10) --- updated-dependencies: - dependency-name: actions/upload-artifact dependency-version: '5' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions - dependency-name: github/codeql-action dependency-version: '4' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions - dependency-name: actions/stale dependency-version: '10' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions ... Signed-off-by: dependabot[bot] <[email protected]> * Check print service name only if remote print service is selected. This fixes the issue that the print dialog is always shown if the local application is selected for printing. * Don't crash MouseSelector, debug GL init (#6334) * beautify (#6342) * Set QSCINTILLA_DLL for MINGW (and not MXECROSS). (#6343) * Warn on offset() with both r and delta specified (#6348) * Fix QFile::Open nodiscard warning with Qt6 6.10.0. (#6364) * Update cmdline help to show Manifold as default (#6350) * Rm unused function. Fixes #6352. (#6353) * Add Color module spec (#6339) * Enables "Ctrl/Cmd-Mouse-wheel zooms text" config option to work (Preliminary fix for #6368) * Entirely removed superfluous EditorInterface::wheelEvent method * Bump Qt to 6.5.7 (#6371) * CircleCI: Update macOS resources to 26.1 on M4 (#6324) * Refactor Discretization Special Usage (#6251) * Tessallation specials $fn$fs$fa now grouped in an object Instead of fn,fs,fa as three doubles, they are passed as an object packaged up with the methods on them. * Repair basic functionality (#6376) * Fix indentation of CC0 template text. (#6375) Also clean up line endings. Add templates/README.md * Fix a number of non-ASCII file name issues (#6136) * Try caching MSYS2 packages Fixes #6381 if it works * Fix `if` in yaml wrong * Don't want `if`s anyway doesn't matter I did them wrong * Update msys2 gh actions caching Make one new cache a day from scratch. Leave note why we don't find the most recent cache. Skip saving since caches are immutable. * Comment updates + build up cache over a month * We can read the determined msys2 directory We're not restoring before the setup, like I originally planned. * Updating script to output list of pacboy packages to workflow Easier than my other approach, actually. Ah, whatever. * Don't change $MSYSTEM in GH Actions * Oops, I need to specify the bash shell I think * Use arrays instead of backslash line termination * Don't fail fast because it fails due to caching a lot well, until we actually get one * std::wstring_convert and std::codecvt_utf8_utf16 is going to be depreciated. Use boost::nowide::narrow (#6356) * Add Unit tests to fix measuring (#6299) * Starting point of release build CI workflow (#6374) * Define executors separately * Added OpenSCAD release build job with create-tarball * Centralize version string management (#6388) * Centralize version string management * Identify build tag in OPENSCAD_VERSION, and correctly extract OPENSCAD_SHORTVERSION * cimg/python: no 'latest' tag. Use 3.14 (#6399) * Disable WIP macOS release build (#6398) * CircleCI Bugfix: Use OPENSCAD_VERSION instead of VERSION (#6400) * create-tarball: Amend files manually since git-archive-all doesn't consider unversioned files (#6401) * cd to /tmp/artifacts before creating shasums (#6402) * Add CMake support + shell helpers for VERSION.txt and COMMIT.txt (#6403) * Ingest VERSION.txt and COMMIT.txt at cmake time * Shell helper for establishing version from VERSION.txt and COMMIT.txt * Snapshot builds now uses the version helpers to unify version management, also deprecated 32-bit windows (#6404) * Bugfix: binary name on macOS needs correct case (#6409) * Parametrize macOS build for release vs. snapshot (#6406) * Parametrize macOS build for release vs. snapshot * Use establish_version.sh and don't require git * Always checkout in the same way; with submodules * Bump actions/checkout from 5 to 6 in the github-actions group (#6418) Bumps the github-actions group with 1 update: [actions/checkout](https://github.com/actions/checkout). Updates `actions/checkout` from 5 to 6 - [Release notes](https://github.com/actions/checkout/releases) - [Changelog](https://github.com/actions/checkout/blob/main/CHANGELOG.md) - [Commits](https://github.com/actions/checkout/compare/v5...v6) --- updated-dependencies: - dependency-name: actions/checkout dependency-version: '6' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * Delete discretizer keyword args PyArg_ParseTupleAndKeywords doesn't allow keyword args it doesn't know about, so we delete them if found. Manually testing the fix. * Fix formatting of "Browser" header (#6433) Markdown rendered is confused by non-breaking space (`U+A0`) - replaced with a regular space. * Make find highlighting case-insensitive * chore: codegen version into manpage (#6448) * chore: codegen version into manpage testing process: - created a VERSION.txt file - ran the CMake build - confirm that the output `build/doc/openscad.1` has the same version as VERSION.txt * remove extra comment * Added job parameters openscad_build_type and openscad_platform (#6449) * Remove deprecated child() and assign(). (#6428) * Remove deprecated child() and assign(). Fixes #5557. Fixes #6425. * Update releases/next.md: child(), assign(), object(), textmetrics(), fontmetrics Remove deprecated child() and assign(). Add object(). Reclassify textmetrics() and fontmetrics() as functions. * More CircleCI cleanup (#6450) * Bump manifold to 3.3.2 and onetbb to 2022.3.0 (#6455) * Added failing test for #6456 (#6459) * Bumped manifold to fix #6235 (#6467) * Temporarily restore exact job names used by circleci-download-artifacts.py (#6469) * CircleCI: Update name of wasm build (#6470) * CircleCI release builds (#6412) * Move version info to .c file (#6463) * chore: bump nix-shell for latest version * convert to flake (for lockfile) * use stable release * remove duplicate lock entry * Add warning to cmake output that ENABLE_CGAL=OFF is a dev-only flag. * Fix font sample rendering in font list window. * Remove obsolete FontListDialog. * Change console font size via mouse wheel event (fixes #6186). This is only changing the current console widget, not the value in the preferences. * Align naming of release jobs, correct naming of tarball (#6484) * Bugfix: Forgot to update dependencies on tarball (#6485) * Bump the github-actions group with 2 updates Bumps the github-actions group with 2 updates: [actions/upload-artifact](https://github.com/actions/upload-artifact) and [actions/cache](https://github.com/actions/cache). Updates `actions/upload-artifact` from 5 to 6 - [Release notes](https://github.com/actions/upload-artifact/releases) - [Commits](https://github.com/actions/upload-artifact/compare/v5...v6) Updates `actions/cache` from 4 to 5 - [Release notes](https://github.com/actions/cache/releases) - [Changelog](https://github.com/actions/cache/blob/main/RELEASES.md) - [Commits](https://github.com/actions/cache/compare/v4...v5) --- updated-dependencies: - dependency-name: actions/upload-artifact dependency-version: '6' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions - dependency-name: actions/cache dependency-version: '5' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions ... Signed-off-by: dependabot[bot] <[email protected]> * Fix premature assignment expiration warning (#6489) * CI: enable sccache (#6473) This avoids doing redundant work by caching compiler results. Because CI does less work, we can run more redundant CI jobs and avoid the mental overhead of manually de-duping jobs. * fix: resurrect LoadShareDesignDialog.h which got lost during merge from OpenSCAD * fix: return type of findModelObject which got inadvertently changed during merge from OpenSCAD * fix: version handling which got broken during merge from OpenSCAD * fix: add missing conv. for editor file path in parseDocument which got lost during OpenSCAD merge * fix: correct variable names in argument parsing for python_cylinder function after OpenSCAD merge * fix: revert test result for use-tests-expected which was inadvertently changed during OpenSCAD merge * fix: update argument parsing for python_cylinder function after OpenSCAD merge * fix: correct discretizer handling in python_cylinder after OpenSCAD merge The OpenSCAD merge introduced CurveDiscretizer to centralize fn/fa/fs parameter handling. However, python_cylinder was creating the discretizer twice: once before argument parsing (which removes fn/fa/fs from kwargs) and again when creating the CylinderNode (from already-modified kwargs). This caused cylinders to ignore fn/fa/fs parameters and render with incorrect facet counts (e.g., 6 sides instead of specified 12). Fixed by: - Removing fn/fa/fs from kwlist (handled by CreateCurveDiscretizer) - Removing unused fn/fa/fs local variables - Using single discretizer instance throughout - Updating format string to match reduced parameter list Resolves incorrect cylinder rendering in hinge and spline tests. Co-Authored-By: Claude Sonnet 4.5 <[email protected]> * fix: add merge-mode-functions flag to gcovr to fix coverage collection After the OpenSCAD merge, gcovr encounters functions in AST.h that appear on multiple lines across different compilation units, causing coverage merging to fail with: GcovrMergeAssertionError: Got function <unknown function> on multiple lines Adding --merge-mode-functions=merge-use-line-0 tells gcovr to use line 0 as the canonical line for functions when merging, resolving the conflict. This fixes the Linux CI coverage step failures while maintaining the same test results (all tests still pass). Co-Authored-By: Claude Sonnet 4.5 <[email protected]> --------- Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: Nicolas Vaagen <[email protected]> Co-authored-by: Torsten Paul <[email protected]> Co-authored-by: Jordan Brown <[email protected]> Co-authored-by: Marius Kintel <[email protected]> Co-authored-by: Ma-XX-oN <[email protected]> Co-authored-by: AaronVerDow <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: coryrc <[email protected]> Co-authored-by: Alex Kent Hajnal <[email protected]> Co-authored-by: Sohler Günther <[email protected]> Co-authored-by: Nicolas Vaagen <[email protected]> Co-authored-by: Glebs Ivanovskis <[email protected]> Co-authored-by: pakhare <[email protected]> Co-authored-by: Flaviu Tamas <[email protected]> Co-authored-by: Guenther Sohler <[email protected]> Co-authored-by: Claude Sonnet 4.5 <[email protected]> * docs: add CLAUDE.md for project guidance and development workf (#288) * feat(python): make add_parameter return value and add feature toggle (#290) - Modified add_parameter() to return the parameter value instead of None - Added feature toggle 'add-parameter-pure-function' to control global variable creation - When feature is disabled (default): returns value AND creates global variable (backward compatible) - When feature is enabled: returns value WITHOUT creating global variable (pure function) - Added two test cases to verify both behaviors This allows gradual deprecation of global variable creation while maintaining backward compatibility. Co-authored-by: Claude Sonnet 4.5 <[email protected]> * chore(master): release 0.8.8 (#289) * fix: evalauting fn for path_extrude again (#287) * fix: evaluating fn for path_extrude again * fix: also wrap works again --------- Co-authored-by: Guenther Sohler <[email protected]> * feat: add children() function * fix(rpm): resolve RPM build failures by forcing static linking (#341) * fix(rpm): force static linking to match Debian package behavior RPM builds were creating shared libraries (libOpenSCADPy.so, libmanifold.so.3, libClipper2.so.1) but not installing them, causing runtime dependency failures. Debian builds statically link these libraries into the main binary, resulting in a working package. Root cause: Fedora's %cmake macro historically sets BUILD_SHARED_LIBS=ON by default, overriding the project's BUILD_SHARED_LIBS=OFF setting. Debian builds don't use this macro and respect the project default. Changes: - Add -DBUILD_SHARED_LIBS=OFF to CMake flags to force static linking - Remove libClipper2.so, libOpenSCADPy.so, libmanifold.so from %files section (these are now statically linked) - Update __requires_exclude to only filter libfive.so (the only bundled shared library that is actually installed) This matches Debian's proven working approach and produces a self-contained binary with only libfive.so as a separate library. Fixes build failures in v0.8.20-v0.8.23 and broken v0.8.19 packages. Co-Authored-By: Claude Sonnet 4.5 <[email protected]> * fix(cmake): use CMAKE_INSTALL_LIBDIR for libfive install path The libfive install rule was using hardcoded "lib" destination, but on 64-bit systems libraries should go to lib64. This caused RPM packaging to fail because the spec file expects libraries in %{_libdir} (which expands to lib64). Changed from: install(FILES ... DESTINATION "lib") To: install(FILES ... DESTINATION ${CMAKE_INSTALL_LIBDIR}) This matches how Debian's rules use CMAKE_INSTALL_RPATH=/usr/lib and ensures libfive.so is installed to the correct architecture- specific library directory. Co-Authored-By: Claude Sonnet 4.5 <[email protected]> --------- Co-authored-by: Claude Sonnet 4.5 <[email protected]> * feat: add installation documentation to website (#342) * feat: add installation documentation to website * feat(web): add installation instructions for GNU Guix * chore(master): release 0.8.24 (#339) * fix(ci): resolve RPM signing passphrase issue (#345) The RPM signing step was failing because the GPG passphrase wasn't being correctly passed through stdin to the rpm/gpg subprocess. The previous approach using --passphrase-fd 0 didn't work in the GitHub Actions container environment. This fix: - Creates a temporary file for the GPG passphrase instead of piping - Configures the %__gpg_sign_cmd RPM macro to use --passphrase-file - Uses rpmsign instead of rpm for clarity - Properly cleans up the passphrase file after signing The passphrase file approach is more reliable across different environments and avoids the stdin redirection issues. Fixes: https://github.com/pythonscad/pythonscad/actions/runs/20981164976 * fix(ci): AppImage libfive dependency, Qt5 support, ARM64 support, and distribution updates (#347) * fix(ci): Fix EOL date for Debian Trixie * ci(debian): stop building packages for Ubuntu 24.10 oracular as it's EOL * ci(debian): fix EOL date for Ubuntu 22.04 LTS jammy * ci(debian): fix EOL date for Ubuntu 24.04 LTS noble * fix(ci): resolve AppImage libfive dependency and add Qt5/ARM64 support This commit fixes the AppImage build and adds multi-architecture support: 1. **Libfive dependency issue**: The AppImage build started failing after commit 50591251c which changed libfive install path from hardcoded "lib" to CMAKE_INSTALL_LIBDIR. Root cause: - On Ubuntu 24.04, CMAKE_INSTALL_LIBDIR defaults to "lib/x86_64-linux-gnu" - libfive.so is now installed to usr/lib/x86_64-linux-gnu/libfive.so - linuxdeploy expects libraries in usr/lib/ and can't find it Error: ``` ERROR: Could not find dependency: libfive.so ERROR: Failed to deploy dependencies for existing files ``` Solution: - Override CMAKE_INSTALL_LIBDIR=lib for AppImage builds - This keeps libfive.so in usr/lib/ where linuxdeploy expects it - Appropriate for AppImage which uses a simplified /usr layout 2. **Add Qt5 support**: Converted the AppImage workflow to a matrix build that creates both Qt5 and Qt6 versions of the AppImage. 3. **Add ARM64 (aarch64) support**: Extended the matrix to build for both x86_64 and aarch64 architectures. Implementation: - Added architecture detection (auto-detect via uname -m or ARCH env var) - Download correct architecture versions of linuxdeploy, appimagetool, and linuxdeploy-plugin-qt - Use QEMU emulation for ARM64 builds on x86_64 runners - Include architecture and Qt version in AppImage filename to avoid naming conflicts Changes: - Added ARCH environment variable to create_appimage.sh - Script auto-detects architecture or uses ARCH environment variable - Workflow builds 4 variants: Qt5/Qt6 × x86_64/aarch64 - Each variant uses the appropriate dependency profile - AppImage filenames include Qt version and architecture: - PythonSCAD-<version>-qt5-x86_64.AppImage - PythonSCAD-<version>-qt5-aarch64.AppImage - PythonSCAD-<version>-qt6-x86_64.AppImage - PythonSCAD-<version>-qt6-aarch64.AppImage Fixes: https://github.com/pythonscad/pythonscad/actions/runs/20981164946 * feat: add ARM64 AppImage support using Docker containers Implements ARM64 (aarch64) AppImage builds using Docker containers, following the same pattern as Debian package builds. This approach uses multi-arch Docker images with QEMU emulation at the container level, allowing proper execution of native ARM64 tools. Changes: - Add arch matrix dimension (x86_64, aarch64) to workflow - Use ubuntu:24.04 Docker container for ARM64 builds - Keep native builds for x86_64 (no container overhead) - Split dependency installation for native vs container contexts - Update artifact naming to include architecture - Pass ARCH environment variable to build script This enables building AppImages for both x86_64 and ARM64 across Qt5 and Qt6, resulting in 4 total AppImage variants. Co-Authored-By: Claude Sonnet 4.5 <[email protected]> * fix(ci): use explicit docker run for ARM64 AppImage builds The previous approach using the container field didn't work because GitHub Actions doesn't support specifying platform/architecture for the container image. This caused Docker to pull the amd64 image even for ARM64 builds, resulting in "cannot execute binary file" errors. Solution: - x86_64 builds: Continue running natively on ubuntu-24.04 runner - aarch64 builds: Use explicit 'docker run --platform linux/arm64' command after setting up QEMU This approach ensures the ARM64 Docker container pulls the correct arm64 image and all tools inside run as native ARM64 binaries. Co-Authored-By: Claude Sonnet 4.5 <[email protected]> * fix(ci): fix ARM64 builds for Debian and RPM packages Both Debian and RPM workflows had the same issue as the AppImage workflow: using the container field pulls amd64 images even for ARM64 matrix entries, resulting in mislabeled packages (amd64 packages with arm64 in filename). Issue discovered: - Debian packages named *_arm64.deb were actually Architecture: amd64 - This occurred because container field doesn't respect platform/arch - QEMU setup never ran due to condition: matrix.qemu && !matrix.docker_image Solution (same approach as AppImage): - amd64 builds: Continue using container field (existing behavior) - arm64 builds: Use explicit 'docker run --platform linux/arm64' - Set up QEMU before ARM64 builds - Wrap all build steps (dependencies, build, signing) in docker run This ensures ARM64 packages are actually built for ARM64 architecture. Testing: - Previous v0.8.24 arm64.deb packages are actually amd64 - New builds will produce genuine ARM64 binaries Co-Authored-By: Claude Sonnet 4.5 <[email protected]> * fix(ci): escape multi-line secrets in ARM64 RPM build script The ARM64 RPM build was failing with 'unexpected EOF while looking for matching quote' error. This was caused by multi-line secret values (GPG keys and passphrases) being substituted into echo commands with double quotes inside a bash -c '...' single-quoted string. When GitHub Actions substitutes a multi-line secret like DEB_SIGNING_KEY into: echo "${{ secrets.DEB_SIGNING_KEY }}" inside bash -c '...', it creates: echo "-----BEGIN PGP... line2 line3..." which breaks bash syntax with unclosed quotes. Solution: Use single quotes around the secrets (with proper escaping for the bash -c context) to preserve them literally: echo '\''...'\'\' This matches the pattern already used successfully in the Debian workflow. Fixes: https://github.com/pythonscad/pythonscad/actions/runs/20997281343 * fix(ci): add docker images for Ubuntu jammy/noble ARM64 builds Ubuntu jammy and noble had docker_image set to null because they use native GitHub runners for amd64 builds. However, ARM64 builds require Docker with the --platform flag, which needs a valid image name. When docker_image was null, the Docker command was malformed: docker run ... ${{ matrix.docker_image }} bash -c ... docker run ... bash -c ... # docker_image=null becomes empty string Docker then interpreted 'bash' as the image name and pulled bash:latest, which doesn't have apt-get, causing the build to fail. Fix: Add explicit docker_image values (ubuntu:jammy, ubuntu:noble) so ARM64 builds work correctly. The amd64 builds still use native runners via the runner field. Fixes: https://github.com/pythonscad/pythonscad/actions/runs/20997278400 * ci: do not build for arm64 at the moment as it is too slow * chore: remove ARM64 from repository website templates ARM64 builds are temporarily disabled due to slow QEMU emulation. Updated HTML templates in update-apt-repo.sh and update-yum-repo.sh to remove arm64/aarch64 from supported architectures list. --------- Co-authored-by: Claude Sonnet 4.5 <[email protected]> * chore(master): release 0.8.25 (#346) * feat(web): improve releases page and add cheat sheet (#348) * feat(ci): implement draft release workflow with auto-publish - Add 'draft: true' to release-please config to create draft releases - Create publish-release workflow that monitors build workflows - Auto-publish when all builds succeed - Keep draft and add status report when builds fail - Ensures downloads page only shows complete releases * fix: fix cheatsheet URL to adapt to new website layout * fix(web): correct HTML structure in cheatsheet - Fix missing closing </div> tags in multiple sections - Change incorrect </section> tags to </div> - Fix <<div typo to <div> - Add missing <div class='grid'> wrappers - Add missing closing </code></pre> tags - Fixes layout issues where sections overlapped or displayed incorrectly * docs: add cheat-sheet page to website * chore(master): release 0.8.26 (#349) * fix(ci): use prerelease instead of draft for release workflow (#351) Problem: Draft releases don't trigger the 'release: [created]' event, so build workflows weren't running after release-please created releases. Solution: Use 'prerelease: true' instead of 'draft: true': - Prereleases DO trigger the release:created event - Build workflows now trigger properly - Publish workflow clears prerelease flag when all builds succeed - Users still won't see incomplete releases (prereleases are hidden) Changes: - .release-please-config.json: Changed draft:true to prerelease:true - .github/workflows/publish-release.yml: Updated to handle prereleases - Check for prerelease flag instead of draft - Clear prerelease flag when publishing - Updated error messages * chore: update from OpenSCAD (#344) * Run once daily. * Update (and run) translation scripts; require bash; enable error options. * Bumped Clipper to 2.0.1 (#6494) * Bumped Clipper to 2.0.1 * Add bugfix for https://github.com/AngusJohnson/Clipper2/pull/1054 --------- Co-authored-by: Torsten Paul <[email protected]> Co-authored-by: Marius Kintel <[email protected]> * chore: remove obsolete Python 3.11 binaries (#353) * chore: remove uncrustify config as it is no longer in use (#355) * chore(master): release 0.8.27 (#356) * fix(ci): fix GPG signing and unify build matrices (#357) * fix(ci): fix GPG signing in non-TTY environments The RPM signing was failing in GitHub Actions with: - "Could not set GPG_TTY to stdin: Inappropriate ioctl for device" - "gpg: signing failed: Bad passphrase" Root causes identified and fixed: 1. Duplicate gpg in macro: `%{__gpg} gpg` expanded to `/usr/bin/gpg gpg` - Fixed by removing the duplicate `gpg` 2. Wrong option order: When `--pinentry-mode loopback` came first, GPG didn't recognize subsequent options like `--batch` - Fixed by putting `--batch` first in the options list 3. Missing gpg-agent configuration: The agent needs `allow-loopback-pinentry` in gpg-agent.conf to allow passphrase injection without a TTY - Fixed by creating gpg-agent.conf before importing keys 4. Using long-form options: Changed `--sign --detach-sign --output` to `-sbo` (the short form used by rpm's default macro) 5. Set GPG_TTY="" to suppress harmless TTY warnings in CI Tested in Docker without TTY to confirm the fix works before applying to the workflow. Co-Authored-By: Claude Opus 4.5 <[email protected]> * refactor(ci): read RPM build matrix from supported-distributions.json - Add Fedora 42, Fedora 43, and Rocky Linux 9 (EL) entries to supported-distributions.json - Refactor build-rpm-packages.yml to use prepare-matrix job that reads from the JSON file, matching the pattern used by build-debian-packages.yml - Both workflows now share a single source of truth for supported distributions and architectures - Update notes section to document supported families Co-Authored-By: Claude Opus 4.5 <[email protected]> * chore: remove arm64 architecture support for Fedora 42 and 43 Drop arm64 from the supported architectures list for Fedora 42 and 43 distributions as build times are currently way too long. --------- Co-authored-by: Claude Opus 4.5 <[email protected]> * fix(web): use theme CSS variables for cheatsheet dark mode support (#358) Replace hardcoded light-mode colors with Material for MkDocs CSS variables so the cheatsheet page correctly adapts to dark mode. Co-authored-by: Claude Opus 4.5 <[email protected]> * chore(master): release 0.8.28 (#359) * fix(ci): import GPG public key to RPM keyring for signature verification (#361) * chore(master): release 0.8.29 (#362) * fix(ci): fix RPM signing for Fedora 43 (RPM 6.x) (#363) RPM 6.0 introduced breaking changes where custom %__gpg_sign_cmd macro overrides no longer work due to parametric macros. This commit adds version-aware signing: - RPM 4.x (Fedora 42, Rocky 9): Use existing approach with %__gpg_sign_cmd - RPM 6.x (Fedora 43): Use gpg-preset-passphrase to cache passphrase in gpg-agent and use %_openpgp_sign_id macro Fixes signing failure on Fedora 43 while maintaining compatibility with older RPM versions. * chore(master): release 0.8.30 (#364) * Update lodepng to 20260102 and add a README * Add to readme where the lexertl source came from Wrote a bash script to confirm this, but it was HEAD of the other repo at the time of the original commit. * Estimate cache size for manifold objects (#6499) * clang-format: Add AllowShortFunctionsOnASingleLine: InlineOnly (#6530) * Correctly preferences after OpenSCADApp has finished initializing (#6533) * Add hidapi README (#6532) Based on comments to https://github.com/openscad/openscad/issues/6513 add a README.md for hidapi like the other external libraries. * Add color list window. * Add support for double-click to insert the color name / color() call. * Ensure newline after HtmlLink message; Skip messages in ErrorList window. * Require "xkcd:" prefix to switch to the XKCD color map. * Suppress log messages of type HtmlLink on command line. * Handle API change between Qt5 and Qt6 (qreal -> float) in QColor::getHslF(). * fix: used other version from regression * fix:udpated tests * fix(ci): fix keygrip extraction for RPM 6.x signing (#366) The previous grep pattern '^\s*sec' expected whitespace before "sec" but GPG output starts with just "sec" (no leading whitespace). This caused the keygrip extraction to fail on Fedora 43 (RPM 6.x). Fixed by using awk to find lines starting with "sec" and then extracting the keygrip from the following line containing "Keygrip". Also added debug output to show the full GPG key details for easier troubleshooting. Co-authored-by: Claude Opus 4.5 <[email protected]> * chore(master): release 0.8.31 (#367) * chore: merge 91 commits from openscad/master (#277) * add correct StartupWMClass to set the correct icon on Gnome I believe this is the XDG way to do this, but I am not sure. * Check for save/discard/cancel on quit. Fixes #3971. (#6311) * Bump CGAL to 6.1.0 (#6318) * Make vector swizzle experimental (#6307) * Don't log to stderr (#6323) * Fully separate zoom levels for each tab. Fixes #4742. (#6320) * Fix/cleanup compiler warnings openscad/openscad#6245 (#6300) * Remove obsolete HTML example converter (#6328) * reformat python tests scripts in pep8 (#6330) * Bump the github-actions group with 3 updates Bumps the github-actions group with 3 updates: [actions/upload-artifact](https://github.com/actions/upload-artifact), [github/codeql-action](https://github.com/github/codeql-action) and [actions/stale](https://github.com/actions/stale). Updates `actions/upload-artifact` from 4 to 5 - [Release notes](https://github.com/actions/upload-artifact/releases) - [Commits](https://github.com/actions/upload-artifact/compare/v4...v5) Updates `github/codeql-action` from 3 to 4 - [Release notes](https://github.com/github/codeql-action/releases) - [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md) - [Commits](https://github.com/github/codeql-action/compare/v3...v4) Updates `actions/stale` from 9 to 10 - [Release notes](https://github.com/actions/stale/releases) - [Changelog](https://github.com/actions/stale/blob/main/CHANGELOG.md) - [Commits](https://github.com/actions/stale/compare/v9...v10) --- updated-dependencies: - dependency-name: actions/upload-artifact dependency-version: '5' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions - dependency-name: github/codeql-action dependency-version: '4' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions - dependency-name: actions/stale dependency-version: '10' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions ... Signed-off-by: dependabot[bot] <[email protected]> * Check print service name only if remote print service is selected. This fixes the issue that the print dialog is always shown if the local application is selected for printing. * Don't crash MouseSelector, debug GL init (#6334) * beautify (#6342) * Set QSCINTILLA_DLL for MINGW (and not MXECROSS). (#6343) * Warn on offset() with both r and delta specified (#6348) * Fix QFile::Open nodiscard warning with Qt6 6.10.0. (#6364) * Update cmdline help to show Manifold as default (#6350) * Rm unused function. Fixes #6352. (#6353) * Add Color module spec (#6339) * Enables "Ctrl/Cmd-Mouse-wheel zooms text" config option to work (Preliminary fix for #6368) * Entirely removed superfluous EditorInterface::wheelEvent method * Bump Qt to 6.5.7 (#6371) * CircleCI: Update macOS resources to 26.1 on M4 (#6324) * Refactor Discretization Special Usage (#6251) * Tessallation specials $fn$fs$fa now grouped in an object Instead of fn,fs,fa as three doubles, they are passed as an object packaged up with the methods on them. * Repair basic functionality (#6376) * Fix indentation of CC0 template text. (#6375) Also clean up line endings. Add templates/README.md * Fix a number of non-ASCII file name issues (#6136) * Try caching MSYS2 packages Fixes #6381 if it works * Fix `if` in yaml wrong * Don't want `if`s anyway doesn't matter I did them wrong * Update msys2 gh actions caching Make one new cache a day from scratch. Leave note why we don't find the most recent cache. Skip saving since caches are immutable. * Comment updates + build up cache over a month * We can read the determined msys2 directory We're not restoring before the setup, like I originally planned. * Updating script to output list of pacboy packages to workflow Easier than my other approach, actually. Ah, whatever. * Don't change $MSYSTEM in GH Actions * Oops, I need to specify the bash shell I think * Use arrays instead of backslash line termination * Don't fail fast because it fails due to caching a lot well, until we actually get one * std::wstring_convert and std::codecvt_utf8_utf16 is going to be depreciated. Use boost::nowide::narrow (#6356) * Add Unit tests to fix measuring (#6299) * Starting point of release build CI workflow (#6374) * Define executors separately * Added OpenSCAD release build job with create-tarball * Centralize version string management (#6388) * Centralize version string management * Identify build tag in OPENSCAD_VERSION, and correctly extract OPENSCAD_SHORTVERSION * cimg/python: no 'latest' tag. Use 3.14 (#6399) * Disable WIP macOS release build (#6398) * CircleCI Bugfix: Use OPENSCAD_VERSION instead of VERSION (#6400) * create-tarball: Amend files manually since git-archive-all doesn't consider unversioned files (#6401) * cd to /tmp/artifacts before creating shasums (#6402) * Add CMake support + shell helpers for VERSION.txt and COMMIT.txt (#6403) * Ingest VERSION.txt and COMMIT.txt at cmake time * Shell helper for establishing version from VERSION.txt and COMMIT.txt * Snapshot builds now uses the version helpers to unify version management, also deprecated 32-bit windows (#6404) * Bugfix: binary name on macOS needs correct case (#6409) * Parametrize macOS build for release vs. snapshot (#6406) * Parametrize macOS build for release vs. snapshot * Use establish_version.sh and don't require git * Always checkout in the same way; with submodules * Bump actions/checkout from 5 to 6 in the github-actions group (#6418) Bumps the github-actions group with 1 update: [actions/checkout](https://github.com/actions/checkout). Updates `actions/checkout` from 5 to 6 - [Release notes](https://github.com/actions/checkout/releases) - [Changelog](https://github.com/actions/checkout/blob/main/CHANGELOG.md) - [Commits](https://github.com/actions/checkout/compare/v5...v6) --- updated-dependencies: - dependency-name: actions/checkout dependency-version: '6' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * Delete discretizer keyword args PyArg_ParseTupleAndKeywords doesn't allow keyword args it doesn't know about, so we delete them if found. Manually testing the fix. * Fix formatting of "Browser" header (#6433) Markdown rendered is confused by non-breaking space (`U+A0`) - replaced with a regular space. * Make find highlighting case-insensitive * chore: codegen version into manpage (#6448) * chore: codegen version into manpage testing process: - created a VERSION.txt file - ran the CMake build - confirm that the output `build/doc/openscad.1` has the same version as VERSION.txt * remove extra comment * Added job parameters openscad_build_type and openscad_platform (#6449) * Remove deprecated child() and assign(). (#6428) * Remove deprecated child() and assign(). Fixes #5557. Fixes #6425. * Update releases/next.md: child(), assign(), object(), textmetrics(), fontmetrics Remove deprecated child() and assign(). Add object(). Reclassify textmetrics() and fontmetrics() as functions. * More CircleCI cleanup (#6450) * Bump manifold to 3.3.2 and onetbb to 2022.3.0 (#6455) * Added failing test for #6456 (#6459) * Bumped manifold to fix #6235 (#6467) * Temporarily restore exact job names used by circleci-download-artifacts.py (#6469) * CircleCI: Update name of wasm build (#6470) * CircleCI release builds (#6412) * Move version info to .c file (#6463) * chore: bump nix-shell for latest version * convert to flake (for lockfile) * use stable release * remove duplicate lock entry * Add warning to cmake output that ENABLE_CGAL=OFF is a dev-only flag. * Fix font sample rendering in font list window. * Remove obsolete FontListDialog. * Change console font size via mouse wheel event (fixes #6186). This is only changing the current console widget, not the value in the preferences. * Align naming of release jobs, correct naming of tarball (#6484) * Bugfix: Forgot to update dependencies on tarball (#6485) * Bump the github-actions group with 2 updates Bumps the github-actions group with 2 updates: [actions/upload-artifact](https://github.com/actions/upload-artifact) and [actions/cache](https://github.com/actions/cache). Updates `actions/upload-artifact` from 5 to 6 - [Release notes](https://github.com/actions/upload-artifact/releases) - [Commits](https://github.com/actions/upload-artifact/compare/v5...v6) Updates `actions/cache` from 4 to 5 - [Release notes](https://github.com/actions/cache/releases) - [Changelog](https://github.com/actions/cache/blob/main/RELEASES.md) - [Commits](https://github.com/actions/cache/compare/v4...v5) --- updated-dependencies: - dependency-name: actions/upload-artifact dependency-version: '6' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions - dependency-name: actions/cache dependency-version: '5' dependency-type: direct:production update-type: version-update:semver-major dependency-group: github-actions ... Signed-off-by: dependabot[bot] <[email protected]> * Fix premature assignment expiration warning (#6489) * CI: enable sccache (#6473) This avoids doing redundant work by caching compiler results. Because CI does less work, we can run more redundant CI jobs and avoid the mental overhead of manually de-duping jobs. * fix: resurrect LoadShareDesignDialog.h which got lost during merge from OpenSCAD * fix: return type of findModelObject which got inadvertently changed during merge from OpenSCAD * fix: version handling which got broken during merge from OpenSCAD * fix: add missing conv. for editor file path in parseDocument which got lost during OpenSCAD merge * fix: correct variable names in argument parsing for python_cylinder function after OpenSCAD merge * fix: revert test result for use-tests-expected which was inadvertently changed during OpenSCAD merge * fix: update argument parsing for python_cylinder function after OpenSCAD merge * fix: correct discretizer handling in python_cylinder after OpenSCAD merge The OpenSCAD merge introduced CurveDiscretizer to centralize fn/fa/fs parameter handling. However, python_cylinder was creating the discretizer twice: once before argument parsing (which removes fn/fa/fs from kwargs) and again when creating the CylinderNode (from already-modified kwargs). This caused cylinders to ignore fn/fa/fs parameters and render with incorrect facet counts (e.g., 6 sides instead of specified 12). Fixed by: - Removing fn/fa/fs from kwlist (handled by CreateCurveDiscretizer) - Removing unused fn/fa/fs local variables - Using single discretizer instance throughout - Updating format string to match reduced parameter list Resolves incorrect cylinder rendering in hinge and spline tests. Co-Authored-By: Claude Sonnet 4.5 <[email protected]> * fix: add merge-mode-functions flag to gcovr to fix coverage collection After the OpenSCAD merge, gcovr encounters functions in AST.h that appear on multiple lines across different compilation units, causing coverage merging to fail with: GcovrMergeAssertionError: Got function <unknown function> on multiple lines Adding --merge-mode-functions=merge-use-line-0 tells gcovr to use line 0 as the canonical line for functions when merging, resolving the conflict. This fixes the Linux CI coverage step failures while maintaining the same test results (all tests still pass). Co-Authored-By: Claude Sonnet 4.5 <[email protected]> --------- Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: Nicolas Vaagen <[email protected]> Co-authored-by: Torsten Paul <[email protected]> Co-authored-by: Jordan Brown <[email protected]> Co-authored-by: Marius Kintel <[email protected]> Co-authored-by: Ma-XX-oN <[email protected]> Co-authored-by: AaronVerDow <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: coryrc <[email protected]> Co-authored-by: Alex Kent Hajnal <[email protected]> Co-authored-by: Sohler Günther <[email protected]> Co-authored-by: Nicolas Vaagen <[email protected]> Co-authored-by: Glebs Ivanovskis <[email protected]> Co-authored-by: pakhare <[email protected]> Co-authored-by: Flaviu Tamas <[email protected]> Co-authored-by: Guenther Sohler <[email protected]> Co-authored-by: Claude Sonnet 4.5 <[email protected]> * feat: multmatrix operates on Vector3d * fix:removed wrong printf * added lasercutr lib * support #2 cuts * feat: start allow to alter existing solid paramters * feat(ci): add native Windows packaging workflow (#371) * feat(ci): add native Windows packaging workflow Add a new GitHub Actions workflow that builds PythonSCAD natively on Windows using MSYS2, packages it with CPack (ZIP + NSIS installer), and runs a sanity test (pythonscad --info) to verify the executable works before release. This addresses issue #360 where MXE cross-compiled builds are missing MinGW runtime DLLs (libgcc_s_seh-1.dll), causing the application to fail to launch on Windows. The native MSYS2 build correctly bundles all required DLLs. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(ci): use workspace-relative path for artifacts Changed from /tmp/out/ (MSYS2-only path) to artifacts/ (workspace-relative) to fix upload-artifact action which runs in native Windows context. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(ci): bundle MSYS2 runtime DLLs in package Add step to use ldd to find and copy all MSYS2 DLL dependencies recursively into the staging directory before packaging. This fixes missing DLL errors: - libgcc_s_seh-1.dll - libcairo-2.dll - libwinpthread-1.dll - libstdc++-6.dll And any other transitive dependencies. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(ci): rewrite DLL bundling to avoid bash subshell issues Changed from recursive function to iterative loop to avoid bash 'local' keyword failures in subshells created by pipes. Uses process substitution instead of pipes and iterates up to 10 times to find all transitive DLL dependencies. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(ci): use CMake's install(RUNTIME_DEPENDENCY_SET) for DLL bundling Instead of manually bundling DLLs with ldd, use CMake 3.21+'s native runtime dependency resolution. This ensures CPack packages include all required MSYS2 runtime DLLs (libgcc_s_seh-1.dll, libstdc++-6.dll, etc.). The RUNTIME_DEPENDENCY_SET feature automatically: - Finds all DLL dependencies of the executable - Excludes Windows system DLLs (api-ms-*, system32, etc.) - Installs them to the package destination This is the proper CMake-based solution rather than shell scripting. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(ci): add Qt plugins for native MSYS2 Windows builds Qt applications require platform plugins (qwindows.dll) to start. Also install style, imageformat, and iconengine plugins for full functionality. Plugins installed: - platforms/qwindows.dll (mandatory) - styles/*.dll (native Windows look) - imageformats/*.dll (image loading) - iconengines/*.dll (SVG icons) Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(ci): add qt.conf for Qt plugin discovery on Windows Qt needs a qt.conf file to know where to find plugins when they're not in the default 'plugins/' subdirectory. This file tells Qt that the plugin directories (platforms/, imageformats/, etc.) are in the same directory as the executable. Contents of qt.conf: [Paths] Plugins = . This should fix the missing toolbar icons issue. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(windows): bundle Qt SVG runtime DLL * fix(windows): use MSYS2 Python stdlib instead of embed package For native MSYS2 Windows builds, the executable is linked against MSYS2's libpython3.13.dll. The RUNTIME_DEPENDENCY_SET copies this DLL to the install directory, but the Python standard library was being installed from python.org's embed package, which has an incompatible structure (python313.dll + python313.zip vs libpython3.13.dll + lib/python3.13/). This mismatch caused Python initialization to fail with: "Could not find platform independent libraries <prefix>" "ERROR: Python 3.13.11 not found. Is it installed?" Fix by installing the MSYS2 Python standard library from the system installation, which matches the ABI of the linked libpython DLL. Co-Authored-By: Claude Opus 4.5 <[email protected]> * ci: disable MXE cross compiled Windows builds They are currently broken and are being replaced by native MSYS2 builds. * feat(windows): improve NSIS installer with start menu, launch option, and multiuser support - Fix NSIS configuration to work for both MXE cross-compile and native MSYS2 builds - Add explicit start menu shortcut creation in PythonSCAD folder and root - Register .scad file association with pythonscad.exe instead of openscad.exe - Add "Run PythonSCAD" checkbox on installer finish page - Add multiuser.nsh for per-user installation support infrastructure - Update URLs and icon references from OpenSCAD to PythonSCAD - Properly clean up shortcuts and file associations on uninstall Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(windows): convert NSIS include paths to native Windows format NSIS requires backslashes for include paths on Windows. Using file(TO_NATIVE_PATH) to convert CMake paths to native format. Also removes unused multiuser.nsh include that had unmet dependencies. Co-Authored-By: Claude Opus 4.5 <[email protected]> * fix(windows): use relative paths for NSIS includes CPack runs NSIS from build/_CPack_Packages/<system>/NSIS/, so use relative paths (../../../) to reference files in the build root. Also copies mingw-file-association.nsh to build dir for consistency. Co-Authored-By: Claude Opus 4.5 <[email protected]> * ci(windows): remove experimental suffix from native build artifacts The MSYS2-based native Windows build is now the primary Windows build, so artifact filenames no longer need the msys2-experimental suffix. Co-Authored-By: Claude Opus 4.5 <[email protected]> * chore: disable pre-major version bump settings in release-please Configure release-please to use standard semver bumping by setting bump-minor-pre-major and bump-patch-for-minor-pre-major to false --------- Co-authored-by: Claude Opus 4.5 <[email protected]> * feat(python): extend add_parameter() with customizer options (#368) * fix: memory bugs in add_parameter Python string handling PyUnicode_AsEncodedString creates temporary bytes objects. PyBytes_AS_STRING returns pointer to internal buffer. Without proper cleanup, this creates dangling pointers causing segfaults. Fixed by copying to std::string before calling Py_DECREF(). Affects: default string values, options lists, and dict labels. * fix: check event type before casting to QKeyEvent Static cast of QEvent to QKeyEvent without checking event type causes heap-use-after-free when event is not actually a key event. Fixed by checking event->type() == QEvent::KeyPress before cast. * fix: heap-use-after-free in add_parameter annotations Assignment::addAnnotations stores raw pointers to Annotation objects. Stack-allocated AnnotationList was being destroyed when function returned, leaving dangling pointers. Fixed by heap-allocating AnnotationList with new, matching the pattern used in CommentParser.cc. * fix(gui): fix memory bugs in customizer parameter handling - ParameterWidget: Fix use-after-free when slider is dragged during F5 re-render - Use deleteLater() for widget cleanup to handle pending Qt events - Keep old parameters alive until widgets are fully deleted - Process events after scheduling deletion to handle in-flight events - MainWindow: Fix crash on window close due to RubberBandManager access - Add isBeingDestroyed flag to guard eventFilter during destruction - Set flag at start of destructor before any cleanup * test: add expected output for add_parameter extended tests * fix(ci): import GPG public key to RPM keyring for signature verification (#361) * Add color list window. * Add support for double-click to insert the color name / color() call. * Suppress log messages of type HtmlLink on command line. * fix: used other version from regression * fix:udpated tests * path points is not writable * sync to master * fix; sync regression * sycn with master * fix: convert face to polyghons when not already * fix: improve lasercut library * fix: face/face cuts working now * fix: create also lasercutmodels from whole shapes instead from faces * fix: returning the final result rather than showing it * make sure, that face2face joints get better slices * feat: implemented export GCODE also pylaser.py got a maximum result size parameter * fix: add missed file * feat: scaffolding gcode export options * feat: added Gcode file format option menu --------- Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: Guenther Sohler <[email protected]> Co-authored-by: nomike <[email protected]> Co-authored-by: Nicolas Vaagen <[email protected]> Co-authored-by: Torsten Paul <[email protected]> Co-authored-by: Jordan Brown <[email protected]> Co-authored-by: Marius Kintel <[email protected]> Co-authored-by: Ma-XX-oN <[email protected]> Co-authored-by: AaronVerDow <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: coryrc <[email protected]> Co-authored-by: Alex Kent Hajnal <[email protected]> Co-authored-by: Nicolas Vaagen <[email protected]> Co-authored-by: Glebs Ivanovskis <[email protected]> Co-authored-by: pakhare <[email protected]> Co-authored-by: Flaviu Tamas <[email protected]> Co-authored-by: Claude Sonnet 4.5 <[email protected]> Co-authored-by: pythonscad-gh-bot <[email protected]>

Fix uninitialized memory read

c7f820f

kintel force-pushed the kintel-error_code branch from cc36ccc to c7f820f Compare January 11, 2026 15:15

kintel mentioned this pull request Jan 11, 2026

Bumped Clipper to 2.0.1 openscad/openscad#6494

Merged

2 tasks

kintel added a commit to openscad/openscad that referenced this pull request Jan 13, 2026

Add bugfix for AngusJohnson/Clipper2#1054

90f948b

kintel added a commit to openscad/openscad that referenced this pull request Jan 13, 2026

Add bugfix for AngusJohnson/Clipper2#1054

442e55a

kintel added a commit to openscad/openscad that referenced this pull request Jan 14, 2026

Bumped Clipper to 2.0.1 (#6494)

a718167

* Bumped Clipper to 2.0.1 * Add bugfix for AngusJohnson/Clipper2#1054

nomike pushed a commit to pythonscad/pythonscad that referenced this pull request Jan 14, 2026

Bumped Clipper to 2.0.1 (openscad#6494)

8d2ebf9

* Bumped Clipper to 2.0.1 * Add bugfix for AngusJohnson/Clipper2#1054

nomike pushed a commit to pythonscad/pythonscad that referenced this pull request Jan 14, 2026

Bumped Clipper to 2.0.1 (openscad#6494)

1b99fa7

* Bumped Clipper to 2.0.1 * Add bugfix for AngusJohnson/Clipper2#1054

nomike pushed a commit to pythonscad/pythonscad that referenced this pull request Jan 15, 2026

Bumped Clipper to 2.0.1 (openscad#6494)

ef0cb48

* Bumped Clipper to 2.0.1 * Add bugfix for AngusJohnson/Clipper2#1054

AngusJohnson closed this Jan 16, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix uninitialized memory read#1054

Fix uninitialized memory read#1054
kintel wants to merge 1 commit intoAngusJohnson:mainfrom
kintel:kintel-error_code

kintel commented Jan 11, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

kintel commented Jan 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

kintel commented Jan 11, 2026 •

edited

Loading