Skip to content

openkal-musl 0.17.0: register it, and it asks nothing of the floor - #447

Merged
Sunrisepeak merged 1 commit into
mainfrom
register-openkal-musl-0-17-0
Sep 20, 2026
Merged

Sunrisepeak merged 1 commit into
mainfrom
register-openkal-musl-0-17-0

Conversation

@Sunrisepeak

Copy link
Copy Markdown
Member

The release

openkal-musl 0.17.0 states what the library does not supply: a top-level [c-abi-absent] table of 24 rows, each naming the shape in which that absence reaches a program (link, enosys, accepted-no-effect), with the package's own CI asserting every row against the 1380 objects it builds. It also moves the implementation pins to openkal-linux 0.15.0 and openkal-windows 0.10.0. No source changed.

min_mcpp does not move, and that is measured

mcpp ignores a top-level table it does not know and refuses an unknown member of a table it does know. The first spelling of this table was [c-abi].absent, and measured against the genuine published 2026.9.18.3 archive — this index's current floor — it made that engine refuse the whole manifest on every target:

error: [c-abi] has no member 'absent'; the members are: builtins, data-model, presents, wchar

Top-level, the same engine builds it. openkal-musl's own CI pins 2026.9.18.3 and is green on 0.17.0, which is the end-to-end form of the same statement. A client stopped below 2026.9.20.1 keeps this package and loses only the note mcpp would attach to a link that failed at one of these names.

All 231 descriptors parse with 2026.9.18.3, so the descriptor grammar is unchanged as well.

Verification

sha256 recomputed from a fresh download of the GLOBAL url — matches
CN asset fetched in full, cmp against the GLOBAL archive — byte-for-byte equal
lua syntax + 5 descriptor lints pass
mcpp xpkg parse, all 231 pass on the pinned 2026.9.18.3

Two things this also fixes

  • The entry lands in all three platform blocks. It went into linux alone first and check_platform_version_parity caught it: the omission would read as "0.17.0 not found" on macOS and Windows while the file plainly contained it.
  • The header note said 0.15.0 was pending and unlisted. It had been released and registered, and so had 0.16.0. A note claiming a published version does not exist reads as missing data rather than a stale sentence, so it is replaced.

0.17.0 states what the library does NOT supply --- a top-level
`[c-abi-absent]` table of 24 rows, each naming the shape in which that absence
reaches a program, with the package's own CI asserting every row against the
objects it builds. It also moves the implementation pins to openkal-linux
0.15.0 and openkal-windows 0.10.0. No source changed.

`min_mcpp` DOES NOT MOVE, and that is measured rather than assumed. mcpp
ignores a top-level table it does not know and refuses an unknown MEMBER of a
table it does know, so the first spelling of that table made every engine
below 2026.9.20.1 refuse the whole manifest on every target. It is top-level
now, and openkal-musl's own CI builds 0.17.0 green against the published
2026.9.18.3 --- this index's current floor. All 231 descriptors parse with
that same engine, so the descriptor grammar is unchanged too.

The hash was recomputed from a fresh download of the GLOBAL url, and the CN
asset was fetched in full and compared byte for byte against it: equal.

Two other things this touches:

  * The entry lands in all three platform blocks. It went into `linux` alone
    first, and `check_platform_version_parity` caught it --- the omission
    would have read as "0.17.0 not found" on macOS and Windows while the file
    plainly contained it.

  * The descriptor's header note said 0.15.0 was pending and unlisted. It had
    been released and registered, and so had 0.16.0. A note claiming a
    published version does not exist reads as missing data rather than as a
    stale sentence, so it is replaced rather than amended.
@Sunrisepeak
Sunrisepeak merged commit b390b4b into main Sep 20, 2026
10 checks passed
@Sunrisepeak
Sunrisepeak deleted the register-openkal-musl-0-17-0 branch September 20, 2026 13:42
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