[09:37] <cjwatson> WTF sampledata
[09:40] <cjwatson> wgrant: There's a sourcepackagepublishinghistory row in sampledata with a distroseries whose distribution doesn't match that of its archive.  Removing the distroseries checks (for non-PPAs) causes the careful-publishing tests to try to republish it.  Should I just make the relevant tests mark that series OBSOLETE and go lalalala, or should I maybe add a distribution condition to getPending*Publications as a guard against this?
[09:40] <cjwatson> I suspect cleaning up that bit of sampledata would be an Aegean-stables exercise
[09:42] <wgrant> cjwatson: I only fixed the publication factories to prevent that situation last week. Fixing a single row shouldn't be too bad.
[09:42] <wgrant> There was a fair bit of test fallout from the code changes, but it wasn't intractable.
[09:42] <wgrant> Let's see which pub it is.
[09:42] <cjwatson> 24 was the one I found
[09:43] <cjwatson> I'm more wondering whether this might be indicating something I need to guard against on prod
[09:44] <wgrant> It is indeed the only one.
[09:44] <wgrant> Hmmmmm.
[09:44] <wgrant> Possibly.
[09:44] <wgrant> I think I only ever did it on staging.
[09:44] <wgrant> But I was playing around with that sort of bug before I had code access, so I forget.
[09:44] <wgrant> Worth checking on staging, I suppose.
[09:44] <wgrant> We should fix the data rather than the code.
[09:44] <wgrant> Since the data fix is quite obvious.
[09:45] <wgrant> (except for sampledata)
[09:45] <cjwatson> Is it?  status = DELETED or something, I guess
[09:45] <wgrant> No, obliterate the rows entirely. They would never have moved out of PENDING.
[09:46] <wgrant> So they are invalid and there must be no artifacts on disk.
[09:47] <wgrant> Hm, so it's mozilla-firefox in ubuntu-test primary.
[09:47] <wgrant> I don't imagine there would be much direct fallout from that.
[09:47] <wgrant> There are four broken SPPHs on prod.
[09:48] <cjwatson> Yeah, I'd just found those on DF
[09:48] <wgrant> One of them is mine, the other three aren't.
[09:48] <wgrant> Oh
[09:48] <wgrant> Two of them are mine, sorry.
[09:48] <wgrant> And the other two are in some other PPA.
[09:49] <wgrant> NCommander's.
[09:49] <wgrant> And there shouldn't be any binaries, since they could have only have been copied from debian/primary, but let's check.
[09:49] <cjwatson> I'm just checking that now on DF
[09:49] <cjwatson> Rather slower query ...
[09:49] <cjwatson> 0 rows
[09:50] <wgrant> BPPH is several times larger, and there's an extra join.
[09:50] <wgrant> Right
[09:50] <wgrant>    id   | archive | distroseries | status | datepublished |        dateremoved
[09:50] <wgrant> --------+---------+--------------+--------+---------------+----------------------------
[09:50] <cjwatson> So these were Debian SPPHs that were erroneously republished in Ubuntu PPAs rather than copied properly?
[09:50] <wgrant>  483272 |      71 |           50 |      4 |               | 2009-01-26 14:03:21.732962
[09:50] <wgrant>  484547 |    2919 |           49 |      4 |               | 2009-06-13 12:47:14.889241
[09:51] <wgrant>  484546 |    2919 |           50 |      4 |               | 2009-06-13 12:47:14.889241
[09:51] <wgrant>  539825 |    7823 |           50 |      4 |               | 2009-07-31 06:13:14.751255
[09:51] <wgrant> Does p-d-r really not restrict by distro?
[09:51] <cjwatson> Anyway, DELETED ones aren't a problem for this
[09:51] <cjwatson> We might want to clean them up anyway, but the publisher isn't going to try to republish them
[09:51] <wgrant> Right, we use the source archive's series in +copy-packages, and the copier used to not validate series.
[09:52] <cjwatson> Oh, that
[09:52] <wgrant> The copier now validates that the series exists in the target, and also the underlying model methods will refuse such madness as of last week.
[09:52] <cjwatson> OK, I guess I fix the sampledata row and run all archivepublisher and soyuz tests or something, then
[09:53] <wgrant> so I think we just DELETE FROM sourcepackagepublishinghistory WHERE id IN (483272, 484547, 484546, 539825); on prod and probably correct the series in sampledata.
[09:53] <wgrant> https://bugs.launchpad.net/launchpad/+bug/320398 ftr
[09:53] <_mup_> Bug #320398: Copy UI uses source distribution series names rather than target distribution series names <lp-soyuz> <package-copies> <phone-rtm> <ppa> <Launchpad itself:Triaged> <https://launchpad.net/bugs/320398>
[09:53] <cjwatson> Yeah, saw that from the phone-rtm list the other day
[09:54] <wgrant> I never thought it would be on me to clean my evil test data up :P
[09:58] <cjwatson> The more I work on Launchpad the more I understand why people hate and despise sampledata
[09:58] <wgrant> Uhuh.
[09:59] <wgrant> It wouldn't be so bad if the sampledata was a vaguely representative sample of anything.
[09:59] <wgrant> But, at least in the Soyuz case, it's mostly invalid crap.
[10:22] <cjwatson> I'm just going to try removing the two broken sampledata rows (one source, one binary).  Let's see who rusts first.
[10:28] <wgrant> Yep.
[17:04] <cjwatson> wgrant: Would you mind re-reviewing https://code.launchpad.net/~cjwatson/launchpad/optimise-publish-a/+merge/226312 ?  I *think* I have it right now, but it took a while to sort out the details.