#11775 Test downgrade kernel on a i686 koji builder
Closed: Fixed 3 months ago by kevin. Opened 4 months ago by kalev.

Would it be possible to temporarily downgrade the kernel on one of the i686 builders? I strongly suspect that webkitgtk build issues on i686 (see https://pagure.io/releng/issue/11952) are somehow due to the builders getting a newer kernel and it would be good to test the theory. I played around with the builds a bit locally and it seems to point in that direction.

Last build that went through was webkitgtk-2.43.4-3.fc40, but I don't know how to check what version of kernel was on the build host back then. Maybe we could pick one of the 6.6.x kernels if it's not clear what version was installed back then?


its in root.log...

DEBUG buildroot.py:802: kernel version == 6.6.9-100.fc38.x86_64

yeah, we can try... I can't right now, but later today or tomorrow possibly.

Ahh, I missed it in root.log - sure, whenever works for you, I don't think there's any particular rush with it. Thanks!

Metadata Update from @phsmoura:
- Issue priority set to: Waiting on Assignee (was: Needs Review)
- Issue tagged with: low-gain, low-trouble, ops

4 months ago

yesterday opencv have problems to build on i686 , https://src.fedoraproject.org/rpms/opencv/c/b645eac1f17dd2c587e0ccc4a9cac1aef5d70c30?branch=rawhide , can we check if it is the kernel please ?

Can you all try and duplicate the problem in staging? If we can see it there we can more easily play with kernel versions, etc...

I sent a build
https://koji.stg.fedoraproject.org/koji/taskinfo?taskID=120050383

to login in stg, we need to know the password of our user on https://accounts.stg.fedoraproject.org/ which may be different of the accounts.fedora.org

i686 build failed

cc1plus: out of memory allocating 11635488 bytes after a total of 86609920 bytes

DEBUG buildroot.py:803: kernel version == 6.7.4-200.fc39.x86_64

Ok, I downgraded the x86 stg builders...

buildvm-x86-02.stg.iad2.fedoraproject.org | CHANGED | rc=0 | (stdout) 6.6.14-200.fc39.x86_64
buildvm-x86-03.stg.iad2.fedoraproject.org | CHANGED | rc=0 | (stdout) 6.6.14-200.fc39.x86_64
buildvm-x86-04.stg.iad2.fedoraproject.org | CHANGED | rc=0 | (stdout) 6.6.14-200.fc39.x86_64
buildvm-x86-05.stg.iad2.fedoraproject.org | CHANGED | rc=0 | (stdout) 6.6.14-200.fc39.x86_64
buildvm-x86-01.stg.iad2.fedoraproject.org | CHANGED | rc=0 | (stdout) 6.6.14-200.fc39.x86_64

can you try now?

i686 built without any debuginfo reduction or memory limitation .

It is the kernel that makes the difference

can you try again now?

I updated them to a 6.8.x rc...

120063306 buildArch (webkitgtk-2.43.4-4.fc41.src.rpm, i686) completed successfully

i686 built without tweaks

ok, thats good... so 6.8.x already appears to have a fix.

So, how urgent is this? We are in beta freeze and I don't want to potentially destablize other builds.
Would it be acceptable to wait until after beta then update to 6.8?

Or do we need this sooner? If so, I suppose I could ask for a freeze break and just update the heavybuilder/i386 builders for now?

Or do we need this sooner? If so, I suppose I could ask for a freeze break and just update the heavybuilder/i386 builders for now?

Your call, but there's a security embargo ending soon and there will of course be no updates for Fedora users on any architecture if it's not possible to build for i686.

Sent a freeze break to the list. Will see what other folks say. :)

for me is urgent :D , we want soname bump jpegXL for Fedora 40

FWIW, I've just managed to rebuild webkitgtk against the new jpegxl in COPR: https://copr.fedorainfracloud.org/coprs/rathann/jpegxl/build/7115556/ .

I upgraded all the x86 builders last night. ;)

So, hopefully we are all back on track now... please re-open or file a new ticket if you see any problems.

Metadata Update from @kevin:
- Issue close_status updated to: Fixed
- Issue status updated to: Closed (was: Open)

3 months ago

on kernel mailing list I reported the problem was gone even with kernel 6.7.x

It is fixed ! I built webkitgtk for F40 on koji [1] koji used the
kernel version == 6.7.6-200.fc39.x86_64 . So maybe it wasn't a
kernel problem , hard to tell , sorry.

Log in to comment on this ticket.

Metadata
Boards 1
ops Status: Backlog