pyapp-kit/psygnal - CodSpeed

psygnal

Blog Docs Changelog

Performance History

Latest Results

docs: Fix and complete numpydoc Parameters sections (#435) * docs: fix and complete numpydoc Parameters sections * docs: fix and complete numpydoc Parameters sections * docs: fix and complete numpydoc Parameters sections * docs: fix and complete numpydoc Parameters sections * docs: fix and complete numpydoc Parameters sections * docs: fix and complete numpydoc Parameters sections * style(pre-commit.ci): auto fixes [...] * Update src/psygnal/containers/_selection.py --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Talley Lambert <talley.lambert@gmail.com>

main

1 day ago

docs: Fix and complete numpydoc Parameters sections (#435) * docs: fix and complete numpydoc Parameters sections * docs: fix and complete numpydoc Parameters sections * docs: fix and complete numpydoc Parameters sections * docs: fix and complete numpydoc Parameters sections * docs: fix and complete numpydoc Parameters sections * docs: fix and complete numpydoc Parameters sections * style(pre-commit.ci): auto fixes [...] * Update src/psygnal/containers/_selection.py --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Talley Lambert <talley.lambert@gmail.com>

main

8 days ago

Update src/psygnal/containers/_selection.py

xuanku-001:docs/numpydoc-params-sections

8 days ago

build: restore mypyc debug symbols; document the actual repro-build fix (#434) #429 disabled all C debug info (`debug_level=0`) to make builds reproducible. That works, but it's a bigger hammer than needed and breaks `-dbgsym` packaging for downstream distros — Debian has to explicitly revert it (pyapp-kit/psygnal#421 (comment)). The actual source of non-determinism is hatch-mypyc's default `tempfile.TemporaryDirectory()` build root, which embeds a random path into the compiled .so's debug info every build. Debian's own packaging already fixes this correctly, without touching debug_level, by pointing `HATCH_MYPYC_BUILD_DIR` at a fixed, pre-created directory (see debian/rules, added in 0.15.1-3) combined with dpkg's default `-ffile-prefix-map`. Verified locally: a fixed build dir alone (default debug_level) produces byte-identical embedded paths across separate builds. Reopens #421 with the analysis; documented inline so this doesn't get re-"fixed" the same way again. Claude-Session: https://claude.ai/code/session_011J3gAnJNwqyWvosDvLwNPE Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>

main

8 days ago

style(pre-commit.ci): auto fixes [...]

xuanku-001:docs/numpydoc-params-sections

8 days ago

build: restore mypyc debug symbols; document the actual repro-build fix (#434) #429 disabled all C debug info (`debug_level=0`) to make builds reproducible. That works, but it's a bigger hammer than needed and breaks `-dbgsym` packaging for downstream distros — Debian has to explicitly revert it (pyapp-kit/psygnal#421 (comment)). The actual source of non-determinism is hatch-mypyc's default `tempfile.TemporaryDirectory()` build root, which embeds a random path into the compiled .so's debug info every build. Debian's own packaging already fixes this correctly, without touching debug_level, by pointing `HATCH_MYPYC_BUILD_DIR` at a fixed, pre-created directory (see debian/rules, added in 0.15.1-3) combined with dpkg's default `-ffile-prefix-map`. Verified locally: a fixed build dir alone (default debug_level) produces byte-identical embedded paths across separate builds. Reopens #421 with the analysis; documented inline so this doesn't get re-"fixed" the same way again. Claude-Session: https://claude.ai/code/session_011J3gAnJNwqyWvosDvLwNPE Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>

main

10 days ago

build: restore mypyc debug symbols; document the actual repro-build fix #429 disabled all C debug info (`debug_level=0`) to make builds reproducible. That works, but it's a bigger hammer than needed and breaks `-dbgsym` packaging for downstream distros — Debian has to explicitly revert it (pyapp-kit/psygnal#421 (comment)). The actual source of non-determinism is hatch-mypyc's default `tempfile.TemporaryDirectory()` build root, which embeds a random path into the compiled .so's debug info every build. Debian's own packaging already fixes this correctly, without touching debug_level, by pointing `HATCH_MYPYC_BUILD_DIR` at a fixed, pre-created directory (see debian/rules, added in 0.15.1-3) combined with dpkg's default `-ffile-prefix-map`. Verified locally: a fixed build dir alone (default debug_level) produces byte-identical embedded paths across separate builds. Reopens #421 with the analysis; documented inline so this doesn't get re-"fixed" the same way again. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011J3gAnJNwqyWvosDvLwNPE

tlambert03:fix-421-reproducible-builds

10 days ago

build: support Python 3.15 (and fix missing locals in compiled EmitLoopError on 3.13+) (#431) * fix: show frame locals in EmitLoopError for compiled builds on 3.13+ On 3.13+ `frame.f_locals` is a FrameLocalsProxy, not a dict. mypyc inserted runtime dict checks (on the typeshed-typed attribute and the `_fmt_locals` param), which raised inside `suppress(Exception)`, so compiled wheels silently dropped the "Local variables" section. On 3.15 typeshed types it as `FrameLocalsProxyType | dict`, which also broke the mypyc build outright. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011J3gAnJNwqyWvosDvLwNPE * build: support Python 3.15 - build cp315 wheels (cibuildwheel v4.2 ships 3.15rc, ABI-stable with final) - skip cibuildwheel tests for cp315 until pydantic-core has cp315 wheels (pydantic 2.14); manylinux has no Rust to build it from source - add 3.15 to the test matrix and classifiers Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011J3gAnJNwqyWvosDvLwNPE * ci: raise asv PR regression factor to 1.5 The 1.15 factor flagged unrelated microbenchmarks (1.25-1.32x) on shared runners in several recent PRs. CodSpeed covers fine-grained regressions. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011J3gAnJNwqyWvosDvLwNPE --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>

main

10 days ago

Latest Branches

0%

docs: Fix and complete numpydoc Parameters sections#435

8 days ago

b261969

xuanku-001:docs/numpydoc-params-sections

0%

build: revert debug_level=0, document the real reproducible-builds fix#434

10 days ago

81f3873

tlambert03:fix-421-reproducible-builds

0%

build: support Python 3.15 (and fix missing locals in compiled EmitLoopError on 3.13+)#431

10 days ago

f98b803

tlambert03:py315

© 2026 CodSpeed Technology

Home Terms Privacy Docs