binNMUs v.s. source-only uploads.
Le Tue, Oct 27, 2009 at 04:26:58PM +0900, Charles Plessy a écrit :
> Le Mon, Oct 26, 2009 at 11:07:19PM +0000, Roger Leigh a écrit :
> > While most developers are conscientious enough to make sure their
> > packages build, one does see enough crap packages that IMO this
> > (minimal) bar should probably be kept.
> I am all for allowing source-only uploads AND requiring from the maintainers a
> proper testing of the packages. Having been able to build the package in a
> clean chroot is anyway not a reliable proof of testing.
By the way, I just realised that binNMUs directly update the binary packages in
Testing, shortcutting the 10 day evaluation period. (See
http://packages.debian.org/squeeze/amd64/r-cran-epibasix/download for example,
where r-cran-epibasix was only binNMUed five days ago).
So being able to do source-only uploads would (will?) have a some advantages
for the maintainers who have to do a serie of rebuildings:
- They can work independantly.
- It is obvious who triggered the rebuild.
- The package history is not lost.
- The package has more testing.
- Builds log are available
(at least here they are invisible: https://buildd.debian.org/pkg.cgi?pkg=r-cran-epibasix).
This is of course already doable with source plus binary uploads, but it would
be nice to be allowed to save the time of doing a local building, since this
the accepted workflow for binNMUs.
Have a nice day,
Tsurumi, Kanagawa, Japan