Skip to content

v1.4.0 hotfix: read the packaged tokenizer as UTF-8 (unblocks the Windows archive) - #747

Merged
JustVugg merged 2 commits into
mainfrom
dev
Aug 1, 2026
Merged

v1.4.0 hotfix: read the packaged tokenizer as UTF-8 (unblocks the Windows archive)#747
JustVugg merged 2 commits into
mainfrom
dev

Conversation

@JustVugg

@JustVugg JustVugg commented Aug 1, 2026

Copy link
Copy Markdown
Owner

The v1.4.0 tag build failed on windows-x86_64, so no archives were published — the release job was skipped because one platform failed, which is the behaviour we want.

UnicodeDecodeError: charmap codec cannot decode byte 0x90 in position 5262
FAIL: packaged k3_tokenizer.py does not parse

The file was fine. The check was wrong: open() without an encoding uses the platform default, which on Windows is cp1252, and cp1252 cannot decode a UTF-8 file. Mine, introduced with the all-engines packaging in #737, and caught on the first tag build that exercised it.

Linux and macOS built and verified cleanly, so the four-engine packaging itself works — 863 KB, all engines present and executable, the launcher resolver satisfied.

After this merges, the v1.4.0 tag moves onto it and the release job re-runs. No version bump: the shipped content is identical, and nothing was ever published under the old tag.

JustVugg and others added 2 commits August 1, 2026 04:50
…NSI page

The v1.4.0 tag build failed on windows-x86_64:

  UnicodeDecodeError: 'charmap' codec can't decode byte 0x90 in position 5262
  FAIL: packaged k3_tokenizer.py does not parse

The file is fine -- valid UTF-8, parses cleanly. The check was wrong: open()
without an encoding uses the platform default, which on Windows is the ANSI code
page (cp1252), and cp1252 cannot decode a UTF-8 file containing non-ASCII bytes.
Reproduced locally: cp1252 raises, utf-8 parses.

Mine, introduced with the all-engines packaging in #737, and caught on the very
first tag build that exercised it -- which is what that verification step is for,
even when the thing it catches is the step itself. No broken archive was
published; the job failed before upload.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
fix(release): read the packaged tokenizer as UTF-8, not the Windows ANSI page
@JustVugg
JustVugg merged commit b085b48 into main Aug 1, 2026
26 checks passed
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.

2 participants