[01:07] <wgrant> cjwatson: Moved the LP meeting back to its traditional time for you, let me know if that isn't ideal.
[08:42] <cjwatson> wgrant: Thanks, it'll do for now and we'll see for the next one in a couple of weeks
[08:42] <cjwatson> Good holiday?
[08:44] <wgrant> Yep, pretty good
[08:44] <wgrant> Nice lunar eclipse on Saturday night, and perfect weather for it.
[08:45] <cjwatson> Oh, shiny.
[08:46] <cjwatson> Any talk of putting together a sprint agenda in last night's meeting?
[08:47] <wgrant> It was discussed, indeed. I need to put my list in writing and then the three of us should compare notes.
[08:47] <cjwatson> Righto.
[09:12] <wgrant> Receiving objects:   0% (4588/4587073), 1.74 MiB | 6.00 KiB/s
[09:12] <wgrant> My connection to Canonical is marvellous tonight.
[09:15] <wgrant> Hmm
[09:15] <wgrant> I wonder if we're exacerbating the LFP problem somehow.
[11:35] <wgrant> cjwatson: Are the ddeb-retriever changes largely done except for that LP MP?
[11:39] <cjwatson> wgrant: Yes
[11:41] <wgrant> cjwatson: Excellent. Do I want to know how long it will take to catch up when a new series is initialised?
[11:42] <cjwatson> wgrant: Not sure yet, can't do much realistic testing until the webservice extension is in place
[11:42] <wgrant> cjwatson: Indeed, but we can guess.
[11:42] <cjwatson> It won't re-download everything, but it'll have a whole bunch of BPPHs to look at
[11:42] <wgrant> eg. does it skip a BPPH early if it sees it has that version already?
[11:43] <cjwatson> It's not as early as it should be yet.
[11:44] <cjwatson> At the moment I'm just going by whether the file exists in the pool to avoid having to stand up a separate database; I want to see how long it would take to get to that point when initing a new series before deciding how much to optimise that.
[11:44] <cjwatson> Oh, blast, did I break turnip ...
[11:44] <cjwatson> [2015-04-08 11:39:18,904: DEBUG/PoolWorker-2] "GET /repo/4/refs HTTP/1.1" 503 None
[11:45] <cjwatson> probably :-/
[11:45] <wgrant> Heh
[11:45] <wgrant> That sounds like the WSGI app failed to start.
[11:45] <wgrant> Which is odd.
[11:45] <cjwatson> VersionConflict: (pygit2 0.22.0 (/usr/lib/python2.7/dist-packages), Requirement.parse('pygit2>=0.21.0,<0.22.0'))
[11:45] <wgrant> If you check the name, version and arch of each BPPH as you see it and compare it to the pool, you should be able to throw away all the duplicates before making any O(bpphs) calls at all.
[11:45] <cjwatson> Oi, turnip, I upgraded you
[11:46] <cjwatson> Comparing to the pool requires getting at the source name.
[11:46] <wgrant> That is a good point.
[11:46] <cjwatson> Well, unless I map it through the already-published Packages.
[11:46] <cjwatson> Which is a possibility.
[11:46] <cjwatson> That's my next optimisation candidate.
[11:47] <wgrant> cjwatson: YOu set requirements.txt, but I didn't see a setup.py change in the diff.
[11:47] <cjwatson> Oh, blah, thanks
[11:47] <wgrant> In fact, the change you made doesn't get used by the charm at all :P
[11:47] <cjwatson> Didn't notice that
[11:47] <wgrant> pygit2-requirements.txt is separate precisely so the charm can avoid it.
[11:49] <wgrant> cjwatson: btw, not sure if you saw the Asana notifications, but the work to get merge diffs from pygit2 turns out to be 0, since merge_commits was added in 0.21.4 and merge_trees in 0.22.1.
[11:50] <wgrant> So the feature was added a couple of weeks after I determined it didn't exist.
[11:55] <cjwatson> Ah, I was also using the wrong mojo manifest, sigh.
[11:55] <cjwatson> wgrant: I didn't notice that, thanks.  Glad that this upgrade work has value :-)
[11:59] <cjwatson> turnip's better again.
[12:01] <wgrant> cjwatson: unset SUDO_USER; Ctrl+R upg; is what I usually do.
[12:01] <wgrant> Which manifest did you try?
[12:15] <cjwatson> The default because I did C-r mojo rather than C-r upg.
[12:15] <cjwatson> And didn't look closely enough :)
[12:19] <wgrant> Heh
[12:19] <wgrant> So just a very long no-op.
[12:20] <cjwatson> Exactly.
[12:20] <cjwatson> So noisy it's hard to notice.
[12:21] <wgrant> The best thing about turnip's ssh frontend is that it doesn't invoke a new Launchpaddy Python process per startup, so checking updatedness is reasonably quick...
[12:21] <wgrant> Rather than spending 4-7 seconds on top of the SSH negotiation.
[23:22] <wgrant> blr: What's this Charm Framework thing?
[23:40] <blr> wgrant: the new name for the services framework
[23:41] <blr> wgrant: the api will most likely not be backwards compatible, we'll have to update our charms.
[23:41] <wgrant> Ah, good.
[23:41] <wgrant> Hopefully the API is less broken!
[23:41] <blr> it does look better