PPA upload fails silently if .changes filename is misformed

from #launchpad
(8:11:21 PM) jturney: I'm having a problem using my PPA; the dput succeeds, but then nothing happens; I've almost certainly made the package wrongly... but I don't get any clue what's wrong
(8:11:53 PM) bigjools: jturney: how long ago did you upload?
(8:12:30 PM) jturney: I tried a few days ago
(8:12:50 PM) jturney: I tried again a couple of times in the past couple of hours
(8:22:20 PM) jturney: How long might I expect to wait for the mail with the .changes file at the moment, then?
(8:47:27 PM) maxb: jturney: If your uploaded .changes is not signed, or is not signed with the right pgp-key, I think you currently get no error email (because in that case launchpad doesn't actually know who did the upload)
(8:52:33 PM) jturney: I'm specifying the signing key id 0x7633C10E, which is the one I have configured on launchpad
(8:54:50 PM) jturney: dput says there is a good signature made with that key when I do the upload
(8:58:41 PM) bigjools: jturney: what maxb said, double check your signatures.
(9:03:09 PM) jturney: http://pastebin.com/m315e9705
(9:03:18 PM) jturney: https://launchpad.net/~jon-turney
(9:10:46 PM) bigjools: jturney: can I see your dput.cf please
(9:12:06 PM) jturney: sure
(9:12:35 PM) jturney: http://pastebin.com/m1d9666a2
(9:13:43 PM) bigjools: cprov, are you around?
(9:19:24 PM) bigjools: mthaddon: still about?
(9:19:47 PM) mthaddon: bigjools: yep
(9:20:25 PM) bigjools: mthaddon: aha - can you help jturney see what happened to his missing upload please? It'll be in the failed upload area on germanium I guess
(9:20:59 PM) mthaddon: bigjools: he doesn't get a log/email response of that?
(9:21:21 PM) bigjools: mthaddon: no, it's some failure mode that prevents email
(9:21:28 PM) bigjools: usually a key problem
(9:38:56 PM) bigjools: jturney: http://launchpadlibrarian.net/21065116/eJdXmBy38Eh5kNV8KxOKurMoN36.txt
(9:39:13 PM) bigjools: thanks for helping mthaddon
(9:39:20 PM) maxb: Hmm - why doesn't launchpad email the owner of the PPA if it can't deduce an identity from the key?
(9:39:33 PM) mthaddon: bigjools: np, thx for talking me through it
(9:39:52 PM) bigjools: maxb: to avoid spamming the innocent
(9:40:19 PM) bigjools: we'll be doing ssh-based uploads in the future which will avoid this problem
(9:42:17 PM) jturney: hmm... ok.... so can I simply remove the ~ppa suffix from the changes file?
(9:43:45 PM) bigjools: you need pidgin-musictracker_0.4.12-2_source.changes
(9:43:47 PM) jturney: oh, or is that telling me <arch> is missing
(9:43:49 PM) jturney: right :-)
(9:44:14 PM) bigjools: this warrants a bug report if you wounld't mind :)
(9:44:21 PM) bigjools: erm *wouldn't*
(9:44:42 PM) jturney: sure
(9:44:47 PM) bigjools: thanks
(9:44:58 PM) jturney: oh, no, thank you :)
(9:45:10 PM) bigjools: we aim to please :)
(9:45:45 PM) bigjools: now, it's Friday night and I need a beer. Goodnight all.

Surely `dput` should identify the broken changesfile name and refuse to upload it. That's pretty much what would happen on a SSH-based which could perform several integrity checks synchronously.

I'm tempted to create a FAQ explaining the current undesired behaviour on the Soyuz-side, invalidate this bug on Soyuz context and target it to `dput` if we all agree.

It's still a valid Soyuz bug because it silently fails rather than returning
the error to the uploader.

So I would say that both Soyuz and dput can do better :)

Why is the filename important at all? It's all about the content, isn't it?

IIRC, I even successfully uploaded things like tmp.changes (c.f. question 51583). This was a big helper because I was able to reuse the same command sequence every time from bash history. Nothing package specific in those commands.

The changesfile name format is important because it's published, and will eventually end up on someone else's disk.

There is a standard filename format tells users a lot about what this file is instead of forcing them to open it and find that out. We don't want, by any means, to disrupt this workflow.

In your case, if you are super-keen on reusing bash-history type `dput *.changes` once and hit UP forever ;)

Celson, thank you for your comment.

I think you are misunderstanding something, but it doesn't really matter anymore. Max Bowsher was kind enough to get me going in the right direction (see question 51583 for further information).

Please, somebody fix this silent discarding of uploads, though.

revision 8044