[03:42] <maxiaojun> is bzr dead
[07:26] <cjwatson> wgrant: Yep, I encountered and fixed one of those sets of ddeb failures, and the other failure was a timing-dependent failure in lib/lp/soyuz/stories/soyuz/xx-person-packages.txt.
[08:26] <cjwatson> wgrant: The "Unembargoed builds" section of lib/lp/soyuz/stories/soyuz/xx-private-builds.txt is deliberately testing a property that my branch makes not true.  Do you think perhaps I ought to be testing just for visibility of the archive owner, rather than of the archive?
[08:28] <cjwatson> Since a private team means the very existence of the archive is confidential, while a private archive whose builds are being copied elsewhere is not quite so confidential.
[08:28] <cjwatson> I'm not sure whether this is test-highlighting-wrong-design or fundamentally-wrong-test.
[08:30] <cjwatson> It is, I suppose, slightly weird that with my branch you can read the +build if you have the link, but /builders refuses to link to it ...
[08:32] <cjwatson> Archive.name is always public, just not necessarily Archive.owner.name.
[08:54] <wgrant> cjwatson: Right, your branch isn't strictly correct, but I felt it was correct enough that we should perhaps not overcomplicate things.
[08:55] <cjwatson> https://code.launchpad.net/~cjwatson/launchpad/testfix-builders-visibility is more correct in that sense, if that's a direction you're happy with.
[08:55] <wgrant> Checking visibility of archive.owner would work, but I'm not sure it's worth the complication.
[08:55] <wgrant> Maybe
[08:55] <wgrant> Really we need to have a definition of build privacy that isn't completely insane
[08:55] <wgrant> But such is launchpad
[08:56] <wgrant> cjwatson: Besides being slightly less obvious, that seems fine.
[08:56] <cjwatson> Arguably this belongs in ViewBinaryPackageBuild or something, but then I'd have had to do something separate for recipes ...
[08:57] <wgrant> It's a pretty unfortunate hack either way, so I'm glad it's restricted to the view :)
[08:57] <cjwatson> And there's possibly an argument that it's the formatter that's trying to poke about inside the build's archive and therefore the formatter should also be the thing that checks whether it's OK to do so
[08:59]  * cjwatson prods the MP button
[09:03] <cjwatson> I suppose an alternative contention is that, if a build's source has been copied to a public archive, you should be able to see the build but not know about the private archive it's building in.  But then links are weird and the build log probably leaks the archive anyway ...
[09:04] <cjwatson> Further alternatively: a saner way to unembargo things would possibly be to copy-and-mangle the build to excise references to its original archive.  (But then that's editing history and there's still the build log problem.)
[09:05] <cjwatson> The definitional problem is kind of inherent to unembargoing, really.
[09:34] <wgrant> cjwatson: Yeah, IMO you should be forced to disclose the existence of your team and archive once you copy anything out of it
[09:34] <wgrant> Anything else is just too messy