Skip to content

docs: update hf-xet - #211

Merged
threexc merged 1 commit into
mainfrom
github-actions/update-doc-for-hf-xet
Aug 7, 2026
Merged

threexc merged 1 commit into
mainfrom
github-actions/update-doc-for-hf-xet

Conversation

@github-actions

@github-actions github-actions Bot commented Aug 7, 2026

Copy link
Copy Markdown
Contributor

Automatically generated PR to document a newly published wheel. Please review it carefully before merging.

If necessary, force-push this branch.

Add version 1.6.0

Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
@github-actions
github-actions Bot requested review from justeph and threexc August 7, 2026 19:02
@threexc
threexc marked this pull request as ready for review August 7, 2026 19:22
@github-actions

github-actions Bot commented Aug 7, 2026 •

Copy link
Copy Markdown
Contributor Author
PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-08-07 20:33 UTC

@threexc
threexc merged commit 94cf881 into main Aug 7, 2026
5 checks passed
@threexc
threexc deleted the github-actions/update-doc-for-hf-xet branch August 7, 2026 20:32
luhenry added a commit that referenced this pull request Oct 10, 2026
All 4 musllinux legs (cp312/313/314/314t) segfault (SIGSEGV, exit 139)
in tests/test_encoding.py::test_http_equiv_content_type_is_honored, a
plain charset-detection test with no recursion; the identical manylinux
legs all pass. The crash site varies between a direct hit inside the
test call (cp314t) and a delayed hit in unrelated pytest teardown on
the other three interpreters, which points to heap corruption in the
native extension rather than a stack-size/recursion problem.

0.4.11, 0.4.12 and 0.4.13 all published musllinux riscv64 wheels for
this package successfully, so this is a 1.0.0-specific regression, not
a musl/riscv64 incompatibility in general. 1.0.0 is a large jump from
0.4.13 and most likely carries a lexbor bump; selectolax's own release
workflow (make_release.yml) builds musllinux wheels but runs no test
suite against them at all, so a musl-only memory bug introduced by such
a bump would ship to every musllinux user unnoticed. Upstream carries a
long history of fresh memory-corruption bugs introduced by lexbor
version bumps (e.g. #211, #217), which is consistent with this.

Drop musllinux from the matrix for now, matching the manylinux-only
pattern used elsewhere in this repo when a libc is a genuine blocker
rather than a missing dependency. Revisit once a newer selectolax
release is available.
luhenry added a commit that referenced this pull request Oct 12, 2026
* selectolax: Add version 1.0.0

Signed-off-by: riseproject-dev[bot] <330740410+riseproject-dev[bot]@users.noreply.github.com>

* selectolax: carry forward riscv64 patches to 1.0.0

Nightly version bump omitted the patches/selectolax/<version> directory for
1.0.0, so the build's unconditional `git apply .../*.patch` step
failed with "No such file or directory" before any build/test ran.
Copied forward the same patch set from 0.4.13 (unchanged since then).

* selectolax: Drop musllinux for 1.0.0

All 4 musllinux legs (cp312/313/314/314t) segfault (SIGSEGV, exit 139)
in tests/test_encoding.py::test_http_equiv_content_type_is_honored, a
plain charset-detection test with no recursion; the identical manylinux
legs all pass. The crash site varies between a direct hit inside the
test call (cp314t) and a delayed hit in unrelated pytest teardown on
the other three interpreters, which points to heap corruption in the
native extension rather than a stack-size/recursion problem.

0.4.11, 0.4.12 and 0.4.13 all published musllinux riscv64 wheels for
this package successfully, so this is a 1.0.0-specific regression, not
a musl/riscv64 incompatibility in general. 1.0.0 is a large jump from
0.4.13 and most likely carries a lexbor bump; selectolax's own release
workflow (make_release.yml) builds musllinux wheels but runs no test
suite against them at all, so a musl-only memory bug introduced by such
a bump would ship to every musllinux user unnoticed. Upstream carries a
long history of fresh memory-corruption bugs introduced by lexbor
version bumps (e.g. #211, #217), which is consistent with this.

Drop musllinux from the matrix for now, matching the manylinux-only
pattern used elsewhere in this repo when a libc is a genuine blocker
rather than a missing dependency. Revisit once a newer selectolax
release is available.

---------

Signed-off-by: riseproject-dev[bot] <330740410+riseproject-dev[bot]@users.noreply.github.com>
Co-authored-by: riseproject-dev[bot] <330740410+riseproject-dev[bot]@users.noreply.github.com>
Co-authored-by: Ludovic Henry <git@ludovic.dev>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant