Conversation
luhenry
added a commit
that referenced
this pull request
Sep 12, 2026
luhenry
force-pushed
the
pygame-ce
branch
4 times, most recently
from
September 13, 2026 04:35
83f37a1 to
6dec89e
Compare
Contributor
|
luhenry
marked this pull request as draft
September 21, 2026 00:36
luhenry
force-pushed
the
pygame-ce
branch
2 times, most recently
from
September 21, 2026 02:00
56f783d to
3b29a17
Compare
Builds SDL2, SDL2_image, SDL2_ttf, SDL2_mixer and portmidi from source at the versions upstream's own manylinux docker_base image uses (Rocky 10 ships no SDL2-devel and EPEL is empty on riscv64), then runs cibuildwheel over the pygame-ce checkout. The mod, fluidsynth-midi, wavpack and opusfile mixer backends and SDL_ttf's harfbuzz shaping are dropped: their dependencies are EPEL-only or have no riscv64 devel package. The matrix stops at cp312-cp314 because numpy, pygame-ce's test dependency, ships riscv64 wheels for those interpreters only; pp311 is dropped for lack of riscv64 PyPy and the free-threaded builds are skipped upstream as well. One patch skips test_arc__correct_drawing: pygame.draw.arc() never returns on riscv64 for some of the rect/angle/width combinations it sweeps, hanging the whole suite rather than failing. The rest of draw_test runs as upstream runs it. Signed-off-by: Ludovic Henry <git@ludovic.dev>
luhenry
marked this pull request as ready for review
September 21, 2026 03:01
luhenry
added a commit
that referenced
this pull request
Sep 21, 2026
…h of cp312-cp314 Two real defects fixed by patches, both hidden by pygame's own test runner: pygame.draw.arc() hangs and segfaults on riscv64 (skipped, localised to two tests out of draw_test's 252), and four tests that need SDL backends this build omits now honour upstream's own PG_DEPS_FROM_SYSTEM marker. The 8-day run of "cancelled" jobs was the workflow's own timeout-minutes: 600, not runner eviction. Left ready for review, not merged.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
pygame-ce2.5.8Compiles pygame-ce's SDL2-based C/Cython extensions via its meson-python build. Upstream publishes no riscv64 wheel.
Mirrors upstream's
build-manylinux.yml.Differs from upstream
harfbuzz-devel, but itslibharfbuzzlinkslibglib-2.0andlibgraphite2, which auditwheel would vendor and whose copyleft sources thegpl_sourcesjob would then have to carry - the same trimbuild-pygame.ymlmadePG_DEPS_FROM_SYSTEM=1- upstream's own marker for a build linked against system SDL dependencies rather than its curated set, which its test suite uses to skip what such a build cannot satisfy (upstream sets it inbuild-ubuntu-sdist.yml,run-ubuntu-checks.yml,dev-check.ymlandbuild-sdl3.yml)python -uon the test command - pygame's runner prints a module's results only once that module finishes, so buffered stdout hides which module stallsMatrix: cp312/cp313/cp314. numpy, the test dependency, ships riscv64 wheels for those interpreters only; pp311 is dropped for lack of riscv64 PyPy, and free-threaded builds are skipped upstream too.
Testing
python -m pygame.tests -v --exclude opengl,music,timing --time_out 300, which runs 2385 tests on each interpreterLicense: Wheel bundles SDL2/SDL2_image/SDL2_ttf/SDL2_mixer/portmidi and several Rocky 10 system libraries (mpg123 and jbigkit are GPL); upstream ships no licence text for any of them, so the build adds it and archives the GPL sources.
Patches
0001-tests-skip-the-arc-rasteriser-tests-on-riscv64.patch-pygame.draw.arc()is unreliable on riscv64:test_arc__correct_drawingnever returns andtest_arc__surface_clipsegfaults, both inside the C rasteriser on a baredraw_arc()call. Because pygame's runner defaults to single-process mode its--time_outnever applies, so the hang took whole jobs down and the segfault took the module's buffered output with it. Running draw_test underpython -X faulthandler -m unittest, then each of its 252 tests in its own interpreter, leaves exactly these two failing and everything else green. riscv64-only: the same command runs 2385 tests in 18s against upstream's own x86_64 2.5.8 wheel (same SDL 2.32.10).0002-tests-extend-PG_DEPS_FROM_SYSTEM-to-opus-MOD-and-HarfBuzz.patch- extends thePG_DEPS_FROM_SYSTEMmarker upstream already honours to three cases it misses: thehouse_lo.opusandsurfonasinewave.xmentries intest_load_object/test_object_namehint, which sit unguarded next to an already-guardedhouse_lo.wv, andtest_font_set_script/test_font_set_direction, which need SDL_ttf built with HarfBuzz. Not architecture-specific - the same four tests fail for anyone building against a system SDL_mixer/SDL_ttf without those optional backends.