Merge ~utkarsh/ubuntu/+source/isc-dhcp:fix-nm-regression into ubuntu/+source/isc-dhcp:ubuntu/devel
Status: | Merged |
---|---|
Approved by: | Christian Ehrhardt |
Approved revision: | ae9021ec47958d48ede072134f1ae5022ad385f7 |
Merged at revision: | ae9021ec47958d48ede072134f1ae5022ad385f7 |
Proposed branch: | ~utkarsh/ubuntu/+source/isc-dhcp:fix-nm-regression |
Merge into: | ubuntu/+source/isc-dhcp:ubuntu/devel |
Diff against target: |
31 lines (+13/-0) 2 files modified
debian/changelog (+7/-0) debian/rules (+6/-0) |
Related bugs: |
Reviewer | Review Type | Date Requested | Status |
---|---|---|---|
Christian Ehrhardt (community) | Needs Fixing | ||
Canonical Server | Pending | ||
git-ubuntu developers | Pending | ||
Review via email: mp+399307@code.launchpad.net |
Description of the change
Hello,
This MP fixes the long due regression in network-manager on ppc64el architecture. TL;DR: it was mostly due to gcc-10 being used. The fix is to not use that in favor of gcc-9.
Long story:
Initially, this patch was introduced in 4.4.1-2.1ubuntu8 and then reverted in 4.4.1-2.1ubuntu9 because of the exactly same regression, which is what's preventing isc-dhcp to migrate to the release pocked now.
Now, whilst this patch was dropped here, it was picked up in Debian (since there we don't really have any gcc-10 problems, et al) and now, after the merge of isc-dhcp's latest version with Debian sid, the patch got re-introduced again. So now, this MP basically disables that patch again and builds this against gcc-9.
I've thoroughly checked the result of regression of network-manager in a ppc64el instance using canonistack.
Before the fix, I could reproduce the regression and after the fix (using my PPA: https:/
8<---8<
autopkgtest [15:15:18]: test urfkill-
autopkgtest [15:15:19]: test urfkill-
urfkill-integration PASS
autopkgtest [15:15:19]: @@@@@@@
wpa-dhclient PASS
nm.py PASS
killswitches-
urfkill-integration PASS
8<---8<
Requesting you to please review and sponsor my upload. Thank you! :)
Hi, gcc_10_ compilation_ issues. patch" really have to go away?
does the patch "Fixed_
isn't it enough to just use gcc-9 and be good without non-applying the patch?
The patch looks very safe except the memcpy, maybe the upstream discussion on this continued and there is a better version of it?
Furthermore could we only use gcc-9 on ppc64el please, and example of that is here /git.launchpad. net/ubuntu/ +source/ qemu/commit/ ?id=952155324e6 f461878a8fbb4f3 3d2a37910a5954
https:/
I we could do both that would be much less deviation from how it is supposed to work.
Finally - we can't gcc-9 build this forever - therefore we need some hope to ever be able to drop this. Due to that IMHO there must be an upstream bug about it (if it exists chime in / if it does not exists file it). Then link it from the commit that you add here. Upstream in this case might be Debian IFF it is only happening when Fixed_gcc_ 10_compilation_ issues. patch is applied.