On 1/3/24 11:48, jpro...@redhat.com wrote:


On 1/3/24 11:23, Vít Ondruch wrote:


Dne 02. 01. 24 v 21:50 Pavel Valena napsal(a):


        My build succeeded everywhere. Nevertheless, there were
        some reports about issues with fibers (e.g.
        https://bugs.ruby-lang.org/issues/20085). So we should
        probably observe and if needed, apply some patch.

        I did a build from my EL9 specific specfile here:
        
https://copr.fedorainfracloud.org/coprs/jackorp/ruby-builds/build/6848355/

        It failed with a bunch of segfaults as well, it seems the
        issue is reproducible on copr infra. The number of failed
        tests is the same as with Pavel's build (103/1871)

        Actually, the hw_info differs for copr and koji. Koji is
        missing `paca pacg` (I guess those are related to the
        mentioned `ASFLAGS=-mbranch-protection=pac-ret`), though
        ssbs is present on both.

        My copr build hw_info.log.gz:
        
https://download.copr.fedorainfracloud.org/results/jackorp/ruby-builds/fedora-rawhide-aarch64/06848355-ruby/hw_info.log.gz
        Passed koji build hw_info.log.gz:
        
https://kojipkgs.fedoraproject.org//work/tasks/6817/111186817/hw_info.log

        Also despite both of the infra are reporting the same
        Vendor ID and Model Name, there are visually half the CPU
        flags missing on Koji compared to copr.


        Hoping that somebody is going to try the patch on my behalf
        (unless I hit the issue myself ;) )


    My second build hit 101 issues, similarly to the previous one.


Hm, I have just done the official build in Koji and again without issues and I have never hit this even doing test in Copr, strange. I am still on hold with the patch.

I am worried that Koji is doing something weird or has some weird setup causing us not to trigger the bug.

I have an RPi 4 lying around, I'll try the build there, see if this issue is affecting that platform, as Fedora supports those arm CPUs.

That was a fun excersize, but (un)fortunately RPi 4 arm CPU does not seem to have PAC support, therefore I cannot reproduce the bug there.

I'd personally include the patch.

(I'd not like upgrading it to Fedora 40 in the future and finding out Ruby programs are segfaulting around Fibers and whatnot.)

Alternatively, this might be worth bringing to fedora-devel, see if someome there is wiser about that patch for ARM platforms.

Jarek
--
_______________________________________________
ruby-sig mailing list -- ruby-sig@lists.fedoraproject.org
To unsubscribe send an email to ruby-sig-le...@lists.fedoraproject.org
Fedora Code of Conduct: 
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: 
https://lists.fedoraproject.org/archives/list/ruby-sig@lists.fedoraproject.org
Do not reply to spam, report it: 
https://pagure.io/fedora-infrastructure/new_issue

Reply via email to