FS#70928 - [glibc] testsuite improvements
Attached to Project:
Arch Linux
Opened by Toolybird (Toolybird) - Tuesday, 18 May 2021, 02:57 GMT
Last edited by freswa (frederik) - Friday, 11 February 2022, 18:51 GMT
Opened by Toolybird (Toolybird) - Tuesday, 18 May 2021, 02:57 GMT
Last edited by freswa (frederik) - Friday, 11 February 2022, 18:51 GMT
|
Details
Since GCC-11, the glibc testsuite bails out early and fails
to show the results summary. How does one know whether the
build is any good or not?
Fix this by cherrypicking upstream patch. With this applied I get almost 100% pass. PKGBUILD patch attached. NOTE: the Arch nspawn container for clean chroot builds blocks a few syscalls needed by the testsuite. This can be overcome by the following specially crafted .nspawn settings file (man systemd.nspawn to find out how to name and where to put the file.) tl;dr version e.g. -> /run/systemd/nspawn/"$USER".nspawn) [Exec] Capability=CAP_IPC_LOCK SystemCallFilter=@clock @pkey There is one other test fail when building on AMD CPU. Upstream patch exists but is not committed yet[1] I note also that the testsuite is not run for the 32-bit glibc build. Maybe nobody cares about the 32-bit libs but this seems less than ideal.. [1]: https://sourceware.org/bugzilla/show_bug.cgi?id=27873 
PKGBUILD-glibc.diff
(1.5 KiB)
|
This task depends upon
Closed by freswa (frederik)
Friday, 11 February 2022, 18:51 GMT
Reason for closing: Fixed
Additional comments about closing: Testsuite runs without errors in [testing]
Friday, 11 February 2022, 18:51 GMT
Reason for closing: Fixed
Additional comments about closing: Testsuite runs without errors in [testing]
I care about lib32 so little that I advocate for lib32 to be removed completed...
Wow, interesting.. That would effectively kill off the multilib repo?
It wouldn't affect me personally but I imagine there would be a few unhappy campers.
Could be time for another one of your famous RFC's? :)