rejects source uploads for existing packages when there are no buildable architectures

[Colin Watson]

I just got this rejection message from Soyuz:

Rejected:
Cannot build any of the architectures requested: arm mipsel

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Format: 1.8
Date: Tue, 14 Oct 2008 14:07:59 +0100
Source: partman-ext2r0
Binary: partman-ext2r0
Architecture: source
Version: 1.15ubuntu1
Distribution: intrepid
Urgency: low
Maintainer: Ubuntu Installer Team <email address hidden>
Changed-By: Colin Watson <email address hidden>
Description:
 partman-ext2r0 - Add to partman support for old ext2 (revision 0) (udeb)
Launchpad-Bugs-Fixed: 256459
Changes:
 partman-ext2r0 (1.15ubuntu1) intrepid; urgency=low
 .
   * check.d/nomountpoint_ext2r0:
     - Make $RET look for a boolean value (thanks, Nicolas Valcárcel;
       LP: #256459).
[...]

Although I can understand why you added this check, this seems wrong for two reasons:

  1) It's already in the archive. We shouldn't have packages in the archive that there is no way to upload.
  2) This impedes our ability to prepare for forthcoming architecture additions. We expect to be adding armel in Ubuntu 9.04, and it would be valuable to upload armel-specific packages in preparation for this. As it happens, partman-ext2r0 is only built for (big-endian) arm at present, but depending on bootloader requirements I could well imagine that it could be useful on armel too. (Thus I would rather not resolve 1) simply by removing packages that we can't currently build.)

[Julian Edwards (julian-edwards)]

That seems a fair rationale to me, unless Celso has anything to add?

[Colin Watson (cjwatson)]

This is proving irritating in the Jaunty sync process. I've blacklisted some packages but would rather be able to remove this blacklisting soon.

[Julian Edwards (julian-edwards)]

Would it be reasonable to whitelist any upload where the package already exists in Launchpad?

Or does it need to go further than that?

[Colin Watson (cjwatson)]

Doesn't make sense to me - just accept it all at the source upload stage, IMO.

[Julian Edwards (julian-edwards)]

Ok, I need to discuss this with Celso then. I'm sure he put this check in to prevent problems further down the pipeline, so removing it could be dangerous without checking the ramifications.

[Colin Watson (cjwatson)]

I can understand why you might want to use this to protect against typos, perhaps. How about explicitly allowing all architectures supported by Debian?

[Celso Providelo (cprov)]

RF 7333