Skia’s +1 Maintenance report from 2025-11-17 to 2025-11-21
My second +1 week, I still like that very much!
Nothing specific was asked of me, and when I announced that I was available on Matrix, nobody reached out to give me specific work, so I’ve been free to follow whatever looked interesting. I wasn’t disappointed: a full transition, a clang crash, some merges, and git-ubuntu improvements!
Work-needed items
- For anyone really well versed into
clang, feel free to go fix that crash
- There is still this
git-ubuntuMP that has one remaining comment about the use of asetthat needs to be addressed. I’ll likely follow-up myself on this.
Sponsorship needed
Nothing here, everything has been sponsored already! Many thanks to @ginggs and @enr0n for supporting me the whole week 
Full logs
I first had a look at some ongoing transitions:
getdpagainstlibgmsh4.14:- MP: Merge into ubuntu/devel : skia/ncr_gmsh : lp:~skia/ubuntu/+source/getdp : Git : Code : getdp package : Ubuntu (thanks @enr0n for the sponsor)
llama.cppFTBFS on ppc64el: that was a nice rabbit hole in the ROCm ecosystem!- It boiled down to
rocblashaving trouble building on ppc64el in the first place becauseclangcrashes: - Upstream
clangbug opened: Frontend dumps core consistently compiling rocblas in Debian/Ubuntu on `ppc64el` · Issue #168727 · llvm/llvm-project · GitHub - Debian bug on
rocblasopened: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1121021
- It boiled down to
Then I’ve had a look at some merges in universe. I was particularly looking at some packages where the TIL was not active anymore:
libstreamvbyte(TIL: William ‘jawn-smith’ Wilson):- I also wanted to drive a full transition, from finish to end, and found this sync to be a good candidate, with only two reverse depends.
- I learned along the way that an Ubuntu delta only changing the list of arches the packages builds on is not worth keeping, thanks @ginggs for your teachings and sponsored sync
- bug: Bug #2131922 “Sync libstreamvbyte 0.5.1-2 (universe) from Debian...” : Bugs : libstreamvbyte package : Ubuntu
- rdepends:
libslow5lib(thanks @enr0n for the sponsor)nanopolish(thanks @ginggs for the sponsor)- Interestingly, this last upload hit a weird LP timing issue, where it built against
libstreamvbyte0instead oflibstreamvbyte1, a good while afterlibslow5libbuilt fine againstlibstreamvbyte1. That sparked some discussions with Launchpad team to find a long term solution to that known and very annoying oddity. See Bug #2132116 “Packages may be removed from -proposed before bein...” : Bugs : Launchpad itself
- Interestingly, this last upload hit a weird LP timing issue, where it built against
htpdate(TIL: Adrien Nader):- Trivial merge, hopefully the last one before we can sync (https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1121116)
- MP: https://code.launchpad.net/~skia/ubuntu/+source/htpdate/+git/htpdate/+merge/496174 (thanks @enr0n for the sponsor)
mmc-utils(TIL: Simon Quigley):- In the end a trivial merge replacing all the delta with a single line that upstream could pick, but it started with me splitting commits of a weird delta and wondering why the build failed due to some
_FORTIFY_SOURCE=3shenanigans. - MP: Merge into debian/sid : skia/merge_1.0-1_0+git20240329.f757f413-1ubuntu1 : lp:~skia/ubuntu/+source/mmc-utils : Git : Code : mmc-utils package : Ubuntu
- Debian bug forwarding the new delta: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1121138
- In the end a trivial merge replacing all the delta with a single line that upstream could pick, but it started with me splitting commits of a weird delta and wondering why the build failed due to some
Along the full week, I spent a bit of time trying to improve git ubuntu submit. Details are in the MP, with comments from @rbasak, among which one still needs to be addressed (thanks for the feedback Robie! ). The main adventure here was actually fixing the build of the snap, because git-ubuntu has a slightly convoluted way to snap itself through its .deb package.