Happy new year everyone…. Here is a POSIX compliance results site…. Don’t know if it is mentioned before …it also includes Haiku …. Tests and their results are both clickable so one can see them.I see some results related to missing headers e.g https://sortix.org/os-test/include/aio/aio_write.haiku.err I suspect that some of them are present in a different directory and maybe the real numbers are higher
3 Likes
We don’t implement AIO, so that’s indeed just missing.
Some of the issues noted here have been fixed since beta5 (e.g. I tested one of the process group ones reported as failed and it succeeded on a nightly.) I fixed some of the other header issues on hrev59287. A few of the other behavioral issues look like ones we should fix, but on the whole I didn’t spot much that looked critical, mostly just "nice-to-have"s.
1 Like
Anyone who is interested can get the latest source code from here sortix / os-test · GitLab and run a ‘make html’ command and have the scores recreated with the latest nightly.
2 Likes
See: POSIX compliance
I usually run the open-posix-testsuite in which Haiku is near successful completion. I’ll try this newer test method.
Haiku R1B5 (hrev59304) x64
- Overall: 89% (3331/3712)
- basic: 94% (332/353)
- include: 91% (2751/2995)
- io: 40% (22/55)
- limits: 95% (38/40)
- malloc: 100% (3/3)
- namespace: 20% (14/70)
- process: 100% (20/20)
- pty: 65% (19/29)
- signal: 100% (31/31)
- stdio: 100% (7/7)
- udp: 86% (94/109)
NOTE: Updated - Jan 10, 2026
2 Likes
@waddlesplash has recently improved POSIX compatibility even more few hours ago…If you try with the next nightly the score should be even better @cocobean
1 Like
Most failing io tests fail due to no F_OFD_SETLK. Will a definition similar to musl make them pass?
This is a new feature in POSIX-2024, merely defining it won’t help, we need to actually implement it in the kernel.
It looks like the BSDs have not implemented it yet. We may wind up waiting till after they do to implement it, unless we have some pressing need for it before then.