On Tue, 16 Sep 2025, Michael Niedermayer via ffmpeg-devel wrote:
Hi all
2 months ago we voted on testing Forgejo vs Gitlab, we picked and tested
Forgejo. And as said in that vote, (and surprisingly, i have not forgotten it)
heres the "after testing" discussion and vote
do we want to keep Forgejo or switch back to the ML workflow
(or something else)
Forgejo was not bad, but I remember some issues:
- When doing individual patch reviews and quoting code the code context
gets messed up, maybe it is references the combined diff?
- Comparing pull request versions does not remove the upstreamed commits
from the comparison, so if a new version is newly rebased, I see
upstream commits as well in the comparison.
- I am not sure what would be the counterpart of doing a last-chance ping
before applying a series in the foregejo system, it was a useful thing
in the ML workflow. Maybe some "auto-apply this in 1 day even if there
are no approvals" checkbox with mail notificaitons could work, and that
would also resolve the issue of self-approval.
- Bumping versions and final commit polishing is a bit tiresome with
foregjo. E.g. it would be nice if I could also edit the commit message,
not just the commit title in the web page...
- Pull requests show up on the mailing list but not force pushes, so
people only reading patches via email won't get notifications of new
versions.
Not sure how many of these are fixable...
F. keep Forgejo as primary forge for patch/git workflow
M. switch back to the ML for patch/git workflow
all GA members can vote, by publically replying here with a
"F." / "Forgejo" vs "M." / "ML"
End time is in 7 days unless teh community wants to extend that.
(Also if results are inconclusive like because a 3rd option emerges
then ill restart this with condorcet on vote.ffmpeg.org)
* If we keep forgejo we will likely transition our issue tracker tickets
into forgejo too, discussing with timo yesterday night indicates that
this likely can be done cleaner and neater than at first expected.
Forgejo issue tracker is seriously lacking features:
- severities
- resolutions
- issue workflow
- votes
- custom keywords for components (e.g. mxf, mpegts)
Simple tags to emulate all these is a hack, e.g. how can you
enforce a single severity or a single resoluton for a resolved issue? Or
do ordering by severity or status?
So this vote being about ML v.s. forgejo should not cause any decision
regarding the issue tracker, that should be another vote if any. Until
forgejo catches up, I'd rather keep trac.
Regards,
Marton
* if we switch back to the ML, we still could have subsystem maintainers
using their own forge
thx
--
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
I have often repented speaking, but never of holding my tongue.
-- Xenocrates
_______________________________________________
ffmpeg-devel mailing list -- ffmpeg-devel@ffmpeg.org
To unsubscribe send an email to ffmpeg-devel-le...@ffmpeg.org
