On 17/03/26 at 01:20 +0200, Peter Pentchev wrote: > On Mon, Mar 16, 2026 at 07:37:28PM +0100, Fabian Grünbichler wrote: > [snip] > > IIRC, in the haskell team, the tag signifies "this package has been uploaded > > and ACCEPTed"... > > FWIW, until very recently I did something similar for all of my packages: > I pushed the commit as soon as I got the "ACCEPTED" message, then I waited > until the next day when my nearest mirror would catch up with unstable, > download the source package, diff it, and if everything was fine, only > then would I push the tag to the Git repo. Right now, this means that > there are a couple of my packages for which debaudit reports that > it could not find the Debian upload tag - because it was not there soon > after the package was accepted into the archive, it was only pushed > many hours later :) > > Now I'm not saying that debaudit is wrong there, I mean, there might be > valid reasons for other people (who run e.g. `debcheckout` soon after > the new version is uploaded) to expect the tag to be there. I'm just > saying this is something that can happen in various workflows. > > These days, if more people use tag2upload, of course that means > the tag will be there way before the package is accepted. However, > in various teams and in various people's personal workflows, there are > many different situations, many quirks, and some of them are not > outright wrong :)
Hi Peter, That workflow (tagging manually after acceptance) should no longer be a problem with debaudit. debaudit now retries to find the relevant tag for 15 days after the initial problem was discovered, so it will eventually catch the tag. Also I recently added https://debaudit.debian.net/git2dsc/regressions/history/sid which will allow to track such (temporary?) regressions in sid. A notification or bug reporting service could easily be built on top of this data (for uploads where the tag used to be available, but isn't for the most recent upload, even after letting a few days pass). Lucas
