You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit d59ee95
Browse filesBrowse the repository at this point in the historyBrowse files
ci: the ruby repo's CI tests itself (stale monorepo job removed) (#819)
* ci: the rspec job tests THIS repo (the stale monorepo-checkout job removed)
The restructure left the rspec job checking out the old monorepo and
running ITS corpus specs -- which now enumerate the maps repo and test
the new Thai encodings under the monorepo's older runtime (the
phasathai failures). A standalone repo tests itself: self-checkout,
full spec suite, SKIP_JS/SKIP_PYTHON per the convention. The
conformance-matrix job moves into this workflow (it ran on the same
triggers).
* fix(WO79): scrub the byte-decode to match the python reference
The byte-decode relabeled invalid sequences as UTF-8; the python
reference decodes with errors=replace (U+FFFD). Ruby String#scrub
matches byte-for-byte. Caught by the C1 conformance matrix: the
invalid bytes crashed first_alternates' split.
* fix(WO79): align the decode cap with the contract — 512 steps
CONFORMANCE.md pins the decoder at a 512-step cap; the runtime
defaults had drifted (translator 100, byt5 256) truncating long
outputs (the fas docs-split rows). The C1 matrix caught it.
0 commit comments