[00:00] <StevenK> wgrant: You have two LFAs with identical SHA1s and filesizes and the old binaries_size was counting them twice (so 18 + 18 = 36); whereas we now weed out duplicates so 18
[00:03] <wgrant> StevenK: LFAs or LFCs?
[00:08] <StevenK> wgrant: The LFA ids were different, I'm not sure if they were FK'd to the same LFC.
[00:14] <wgrant> StevenK: We were just distincting on LFC before, weren't we?
[00:14] <wgrant> I don't quite remember.
[00:14] <wgrant> But anyway, yes, it seems likely the test is buggy
[00:16] <StevenK> wgrant: Both LFA and LFC are distinct.
[00:16] <wgrant> StevenK: Now or before?
[00:16] <wgrant> Now they are
[00:17] <StevenK> wgrant: Sorry, in the test, the two publications have distinct LFAs and LFCs.
[00:17] <wgrant> But the LFCs are identical?
[00:17] <StevenK> wgrant: http://pastebin.ubuntu.com/5629884/
[00:18] <wgrant> Right, so normally librarian-gc would have coalesced them
[00:18] <wgrant> The test is wrong
[00:18] <StevenK> So I have no idea how it was coming up with 54.
[00:18] <StevenK> Since that's mentioned in a comment
[00:18] <StevenK> 36 I get, but it's still wrong
[00:21] <StevenK> wgrant: http://pastebin.ubuntu.com/5629892/
[00:22] <wgrant> StevenK: Sounds good
[00:23] <StevenK> wgrant: That file was a bit crazy
[00:23] <StevenK> Less so now
[01:59] <StevenK> wgrant: The source query takes ~ 400 ms, the binary query is more like 1 second
[01:59] <StevenK> (For ubuntu-langpack)
[02:24] <wgrant> StevenK: Sounds OK
[02:24] <wgrant> StevenK: Tried on some Kubuntu PPAs?
[02:44] <StevenK> wgrant: You assume I know where some of them live.
[02:48] <wgrant> StevenK: ~kubuntu-ppa
[02:48] <wgrant> ppa:kubuntu-ppa/backports is quick once the cache is hot
[02:49] <wgrant> When cold, the binary query took 11s. 6s of that was just finding the BPPHs, so inlining the size would have only saved 50%, even though it would also have been wrong.
[02:49] <StevenK> I thought about it a bit more
[02:49] <StevenK> We could solve the dupe issue by also denorming a checksum
[02:50] <wgrant>    ->  HashAggregate  (cost=189318.78..189411.75 rows=9297 width=93) (actual time=11545.856..11547.182 rows=1719 loops=1)
[02:50] <wgrant>          ->  Nested Loop  (cost=737.92..189225.81 rows=9297 width=93) (actual time=3700.384..11534.203 rows=2587 loops=1)
[02:50] <wgrant> Not very effective
[02:50] <wgrant> ~30% decrease in work
[02:50] <StevenK> So the queries still need a little tuning, or there isn't much we can do?
[02:51] <StevenK> Without a massive amount of design, that is.
[02:56] <wgrant> There's not much we can do.
[03:50] <StevenK> wgrant: https://code.launchpad.net/~stevenk/launchpad/back-to-agpl/+merge/154251
[03:53] <wgrant> StevenK: most of that is the former lazr-js?
[03:54] <wgrant> Yes
[03:55] <wgrant> r=me
[06:15] <StevenK> Bleh, I do not understand these JS failures
[06:22] <wgrant> StevenK: Have you tried it locally?
[06:22] <wgrant> It's possible that it makes the files too big or something
[06:22] <wgrant> We've had that problem before, though it's meant to be fixed
[06:23] <wgrant> Ah
[06:23] <wgrant> StevenK: You have some syntax issues
[06:23] <wgrant> [06:23] <wgrant> [06:23] <wgrant> [06:23] <wgrant> [06:24] <wgrant> [06:45] <StevenK> wgrant: I've fixed all of those
[06:48] <StevenK> http://pastebin.ubuntu.com/5630415/
[06:49] <wgrant> StevenK: Does it work?
[06:49] <StevenK> wgrant: It only fixes 3 of the 7 failures
[06:50] <wgrant> StevenK: Have you examined the common files included by those broken tests?
[06:51] <StevenK> Yes, they look fine to me
[08:57] <adeuring> good morning
[10:19] <stub_> http://pgfoundry.org/projects/emaj/
[10:19] <stub_> If that is what I think it is, it might give us an easier way to reset db state after running a test.
[10:20] <stub_> Not sure if it would be a win though from test speed perspective (faster doing the db reset, but inserts, updates and deletes will be slower)
[16:09] <czajkowski> aloha
[16:09] <czajkowski> adeuring: where is the best place to point a canonical person at to fix something in LP
[16:10] <adeuring> czajkowski: the most up to date knowledge have wgrant and StevenK. But i can help perhaps too
[16:11] <adeuring> or other people who retired fromLP development.
[16:11] <czajkowski> mfisch: what area do you want to try and fix I guess would be the best start
[16:11] <czajkowski> adeuring: aye just timezones makes it kinda hard
[16:12] <adeuring> czajkowski: yeah, that's why i sugegsted "retired people" too ;)
[16:12] <mfisch> hey czajkowski, I ran into a problem yesterday where I changed permissions on a script by accident, it was lost in the review because it's hidden in one line in all black text and looks like a "file header", all the reviewers missed it too
[16:12] <mfisch> I was thinking this morning that permissions changes could be called out with colorization as well
[16:13] <mfisch> I don't know anything about the LP code base, but I thought something like: foo.sh <red>-x</red> <green>+x</green>
[16:13] <mfisch> that seemed to be simple enough to call attention
[16:13] <adeuring> mfisch: so, that's about reviews? abentley might be able to help
[16:14] <mfisch> yeah, about reviews
[16:15] <mfisch> I'd be happy to try and fix it myself if I can get a kick-start in the right direction
[16:16] <adeuring> mfisch: a very vague pointer: the directory lib/lp/code/browser
[16:16] <adeuring> in the LP source code.
[16:16] <adeuring> but you'd need to read how to build LP locally on dev.launchpad.net first
[16:17] <mfisch> ok
[16:17] <mfisch> I'll grab the code and poke around and save this for 10% time
[16:18] <adeuring> mfisch: there are also some procedures, like having a pre-imp conversation (what we are beginning now, but I am not sure how far we can get, since i am not very familiar with this part of LP)
[16:20] <mfisch> adeuring: ok, let me look into whether this is feasible too and then I'll talk to abentley too
[16:21] <cjwatson> wouldn't that be in loggerhead?
[16:21] <mfisch> still pulling the code...
[16:22] <cjwatson> oh, ok, not for the review interface I guess
[16:23] <abentley> mfisch: I believe you want to look at FormattersAPI.format_diff in lib/lp/app/browser/stringformatter.py
[16:23] <cjwatson> general advice: reproducing it in the test suite early is generally much more fun than trying to run a local LP instance.
[16:23] <cjwatson> IME.
[16:24] <cjwatson> and you'll need a test anyway
[16:24] <mfisch> okay good advice
[16:24] <mfisch> at this rate I'll have the code by tomorrow
[16:27]  * abentley is somewhat irked that the code parses the diff manually instead of using bzrlib.patches.parse_patches.