Re: Build logs from local builds
Le Wed, Nov 11, 2009 at 10:45:57AM +0100, Stefano Zacchiroli a écrit :
> Personally, I think that the extreme trade-off of making source upload
> the default (which seems to be what you are arguing for) would be too
> risky in term of degraded package quality. Look for the "FTBFS" string
> in the current RC bug list, do you think the number of occurrences we
> have now would decrease implementing such a proposal?
my gut feeling (but maybe I start to sound like a broken disk) is that most
“FTBFS” and RC bugs that stay unfixed are more the signature of abandonned
packages than sloppy maintainers.
Source uploads are a missing tool in our box, and it is a pity that we do not
allow them because we fear that a significant number of DDs would misuse it.
(And let's remember, the average DD is a MIA DD). Source uploads would be a
nice alternative to binNMUs for mid-scale rebuild efforts, where binNMU is too
unclean (no trace in the package history, no 10-day waiting period in Unstable,
and difficult to track because half of our infrastructure seem to ignore them
(see http://packages.debian.org/sid/r-cran-epibasix for instance).
And for the Lintian checks, not only we could live without them for years, but
also as it was written many times, they would give their best if placed at the
output of the buildd network, because of all the race conditions that arise
from the fact that the ‘clean chroots’ on which binary packages are locally
built will not be the same in the buildds when they process the packages a few
days later (even on amd64 this month, see
Have a nice day,
Tsurumi, Kanagawa, Japan