Re: Builds for hurd-i386 using debian-ports.org
On 04/20/2012 01:47 PM, Svante Signell wrote:
> On Fri, 2012-04-20 at 12:45 +0200, Samuel Thibault wrote:
>> Ansgar Burchardt, le Fri 20 Apr 2012 12:10:33 +0200, a écrit :
>>> I do not find it acceptable for a port in the main archive to rely on
>>> external packages, esp. should it eventually be released with wheezy.
>> It is not acceptable either that patches take so long to get uploaded
>> in the main archive. Not using debian-ports would make a new port
>> completely impossible. kfreebsd also did that. That's the only way to go
>> forward, otherwise there is always something broken that prevents from
>> actually making progress.
I am not quite sure, but didn't kfreebsd stop relying on debian-ports
when they moved to the main archive? I only followed the more recent
addition of the new Linux ports.
> If the Debian maintainers were more responsive to bug reports wrt
> GNU/Hurd the debian-ports archive would almost empty by now...
> The problems are not only due to the porters. Currently there are 39
> bugs with patches, 5 forwarded and 6 pending upload, all with severity
> important in the BTS.
If maintainers do not react, couldn't you just go for NMUs instead of
uploading changes to debian-ports?
As you plan to stop using debian-ports, can you estimate when this can
be done? If we plan to rebuilt all binaries after this, it should be
done fairly sure (ie. even before an eventual migration to testing).
We would also need some way to keep track of which binaries still need
to be built in a clean environment. For armhf and s390x this was done
by doing the initial import with a special key and then making sure all
packages signed with this got rebuilt, but we cannot do the same for
hurd-i386. A first idea would be to note when you dropped debian-ports
from the buildds and then making sure all binaries uploaded before that
are rebuilt, but this misses at least arch:all packages.