Control: retitle -1 golang-github-microsoft-didx509go and
golang-github-microsoft-go-winio in bad state on dgit-repos
Control: clone -1 -2
Control: retitle -2 Erroneous taints when archive doesn't process queue for too
long
Control: clone -1 -3
Control: retitle -3 If repo doesn't exist, d-r-s can try to access
'$repos/_empty'
Control: severity -3 minor
Bug gardening:
Ian Jackson writes ("Re: Bug#1127673: dgit-repos-server: fatal: detected
dubious ownership in repository at '$repos/_empty'"):
> Ian Jackson writes ("Bug#1127673: dgit-repos-server: fatal: detected dubious
> ownership in repository at '$repos/_empty'"):
> ...
> > The .changes was generated "Sat, 09 Aug 2025 08:25:35 -0400" ie
> > 2025-08-09T12:25:35Z. The .dsc contains this
> ...
> > It appears that at 2025-08-09T16:37:27Z, the dgit-repos service
> > decided that your package had been REJECTed. There is an entry in our
> > taint database with that timestamp, for the commit names above, saying:
>
> I have just noticed that 2025-08-09 was the date of the trixie
> release. IIRC the whole legacy Debian Archive system was suspended
> for several days for release admin. I think that would explain
> everything.
...
> AFAIAA this kind of breakage hasn't happened before, but we should
> perhaqps consider a way to avoid it happening again. I'm not sure
> what that would look like.
The desire to somehow avoid this happening again (likely, in two
years' time) is going to be bug #-2
Erroneous taints when archive doesn't process queue for too long
> Subject to confirmation from Reinhard I propose to manually fix this
> up on the server, by restoring the two repositories (from "_garbage")
> and by removing the taint indications from the database.
The broken state on the server for these two packages is this bug
which I am therefore retitling
#1127673
golang-github-microsoft-didx509go and golang-github-microsoft-go-winio in
bad state on dgit-repos
That the system is trying to access a bogus repository path will be #-3
If repo doesn't exist, d-r-s can try to access '$repos/_empty'
Which I'm setting to "minor" because it doesn't seem to happen except
in quite strange situations where things have already gone wrong,
and the consequence is it bombing out rather than making anything worse.
Ian.
