On 17/04/2026 15:20, Pádraig Brady wrote:
On 17/04/2026 10:17, Bruno Haible wrote:
Pádraig Brady wrote:
coreutils:
tests: avoid a failure on musl
build: fix build failure on AIX
tests: avoid false failure on OpenBSD
tests: avoid false failure with hidden /proc on Guix
tests: avoid failure on Manjaro Linux
cksum: fix --length validation on 32 bit platforms
tests: fix false failure due to ulimit usage
tests: avoid failure on older Centos 7
tests: fix false failure on Centos 7
Taken together, this snapshot or the previous one passes all tests on:
* GNU/Linux
- CentOS 6, 7, 8
- Slackware 14, 15
- Manjaro 23
- Arch Linux 24.04
* Alpine Linux 3.18
* FreeBSD 14.0
* OpenBSD 7.8
* Solaris 11 OpenIndiana
Also all tests pass on:
Fedora 43 amd64
Debian 11 amd64
Alpine 3.22 riscv64
MacOSX 26 arm64
Alma 9.7 ppc64le
Debian 14 ppc64
Solaris 11.4 sparc
All coreutils tests pass and no new gnulib failures on:
CheriBSD 15 arm64
3 coreutils test failures on older mips (cfarm230).
1 was due to timeout(1) not handling SIGRTMAX==127.
I was tempted to work around that in timeout like:
- else if (WIFSIGNALED (status))
+ /* The status==WTERMSIG(status) check is to handle an ambiguity
+ on mips where there are 128 signals, thus SIGRTMAX overlaps
+ with the WIFSTOPPED bit pattern, thus causing WIFSIGNALED()
+ to return false. */
+ else if (WIFSIGNALED (status) || status == WTERMSIG (status))
However I see there has been fix in glibc 2.44 for it:
https://sourceware.org/git/?p=glibc.git;a=commit;h=c19cdbf6
So I'll refrain from handling in timeout for now.
The other 2 were due to strace 4.26 (2018) giving
"Failed to tamper with ..." errors
and also having tampered calls always returning with ENOSYS (Function not
implemented)
Again due to the age I'll leave this for now.
cheers,
Padraig
