Hi Niklas

On Thu, Apr 02, 2026 at 06:58:48PM +0200, Niklas Haas via ffmpeg-devel wrote:
> Hi all,
> 
> I started a vote about the best way to switch FFmpeg's internal copy of 
> checkasm over to upstream libcheckasm. If you are a member of the GA, you 
> should have received a vote link. If not, please contact me.
> 
> I have tentatively set the end date to April 8th, which I believe gives 
> enough time to read the summary and come to a conclusion. If you need more 
> time to decide, please reply to this e-mail.
> 
> For context, see the discussion at 
> https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/22546

As said elsewhere, if there is to be a vote, it should be done properly.

A vote should be announced before it starts, so people have time to discuss the 
question and suggest options.
Anton and Thilo have both done such pre-vote announcements in the past.

More generally, voting should be a last resort after discussion has failed to 
reach consensus.
If discussion has already brought us close to consensus, then starting a vote 
risks creating unnecessary conflict rather than resolving a real disagreement.

Specifically, I think this vote is missing at least one important option:
1. one flat check-in per upstream commit
   no merge commits, full author attribution, and straightforward use for users 
and developers

For completeness, this could also be an option, though I do not insist on it if 
there is concern about having too many choices:
2. further discussion / do not merge checkasm at this time; keep using the 
current FFmpeg checkasm code for now

Also, the option that corresponds most closely to the current PR should be 
identified explicitly as such.
My impression is that option B is closest to the current consensus, in case 
someone wants to vote for that.
But I can live with a range of solutions, provided that the result is not 
submodules.

About submodules:
Multiple people, including myself, have objected to them.
This is not just a checkasm integration detail. Choosing submodules would be a 
broader structural decision affecting FFmpeg as a whole, and voters should 
understand that clearly.

Among other things:
* submodules complicate workflows such as bisection
* submodules create extra friction for users and developers
* submodules are not a good fit for a scalable development model of the kind 
FFmpeg may want in the future
* as release manager, I will need to replace them one way or another in release 
branches
  this would add all the mess from submodules in git master and another 
solution in release branches.

So I think the consequences of submodules should be presented very clearly if 
that option remains on the ballot.

thx

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Breaking DRM is a little like attempting to break through a door even
though the window is wide open and the only thing in the house is a bunch
of things you dont want and which you would get tomorrow for free anyway

Attachment: signature.asc
Description: PGP signature

_______________________________________________
ffmpeg-devel mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to