[pkg-go] Bug#915907: rclone: autopkgtest regression: missing versioned dependency?

[Paul Gevers]

Hi Tobi,

On 16-12-2018 14:15, Dr. Tobias Quathamer wrote:
> I've now uploaded a new version of rclone which passes (as expected)
> with the current golang-github-okzk-sdnotify package from sid.

Great.

> However, the debci test now fails in testing, because the
> golang-github-okzk-sdnotify package has already migrated to testing,
> while rclone did not. Therefore, the older version of rclone in testing
> now fails to build with the newer okzk-package.

Ack.

> The testing migration time (of okzk) has now been increased by 34 days
> due to the failing test. However, in my opinion, the test result is
> bogus, because the rclone package in sid builds fine with okzk in sid.

I don't agree. autopkgtest is about regressions in testing, not in sid.

> So both packages could migrate together without the increased delay. Or
> am I missing something here?

Yes, a versioned breaks on rclone in okzk or a versioned test dependency
on okzk in rclone. If you add the former, the migration software will
know that it has to tests rclone from unstable instead of rclone from
testing when it runs the autopkgtest for okzk. If you add the latter,
okzk from unstable will be used when testing the rclone in testing, okzk
will be added to the list of triggers, and the result will be used for
okzk of that same version as well. You can pick what describes the
situation best.

> Could you do anything about the increased testing migration? Or can I do
> something about that?

See above.

Paul

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://alioth-lists.debian.net/pipermail/pkg-go-maintainers/attachments/20181216/9ca66650/attachment.sig>