Camelia, the Perl 6 bug

IRC log for #darcs, 2013-03-06

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:30 mizu_no_oto joined #darcs
01:52 mizu_no_oto joined #darcs
02:13 mizu_no_oto joined #darcs
02:22 intripoon joined #darcs
02:56 favonia joined #darcs
04:18 favonia joined #darcs
04:46 mizu_no_oto joined #darcs
04:57 preflex_ joined #darcs
06:05 loochao joined #darcs
06:15 loochao left #darcs
07:03 mornfall joined #darcs
07:28 lelit joined #darcs
08:01 schlaftier joined #darcs
08:20 amgarchIn9 joined #darcs
08:28 mulander joined #darcs
08:51 raichoo joined #darcs
10:01 donri joined #darcs
11:42 lelit joined #darcs
11:42 lelit` joined #darcs
11:47 owst joined #darcs
12:05 iago joined #darcs
12:47 favonia joined #darcs
13:10 mizu_no_oto joined #darcs
13:53 lelit`` joined #darcs
13:59 mizu_no_oto joined #darcs
14:02 amgarchIn9 joined #darcs
14:03 lelit` joined #darcs
14:17 mizu_no_oto joined #darcs
15:57 favonia joined #darcs
16:44 favonia joined #darcs
17:19 Lemmih joined #darcs
17:45 whaletechno joined #darcs
17:57 raichoo joined #darcs
18:02 amgarchIn9 joined #darcs
18:12 snorble_ joined #darcs
19:18 favonia joined #darcs
19:21 schlaftier joined #darcs
19:37 lelit joined #darcs
19:38 Kindaro joined #darcs
20:06 amgarchIn9 joined #darcs
20:59 mizu_no_oto joined #darcs
21:06 favonia joined #darcs
21:07 Kindaro joined #darcs
21:23 carter joined #darcs
21:46 amgarchIn9 how do you create repos on h.d.n? Through the web interface?
21:47 amgarchIn9 are there repository format incompatibilities with the initial push?
21:49 favonia joined #darcs
21:50 sm hi amgarchIn9
21:50 Heffalump amgarchIn9: it won't take darcs-1 format
21:50 sm you can create via web, or ssh (cf http://hub.darcs.net/simon/​hub.darcs.net-docs/FAQ.md)
21:53 mizu_no_oto joined #darcs
21:54 sm amgarchIn9 is helping me troubleshoot a stray lock file
21:55 sm amgarchIn9: so when you push "too large" patches, eg an initial commit of many files, darcs can use too much memory - on darcs hub server, and possibly on your machine too
21:55 sm or so it sounds from your report
21:56 sm do you remember doing a big push like that ? if not maybe it's repo format related
21:59 amgarchIn9 that repo is 1800 pathces 18 kloc pristine. Let me try pushing from ubuntu lts
22:02 favonia joined #darcs
22:03 amgarchIn9 http://pastebin.com/Ery5nUb6 darcs: fd:11: hPutBuf: resource vanished (Broken pipe)
22:03 amgarchIn9 local darcs was at 3.6G RSS according to top
22:04 amgarchIn9 2.7.99.1 (release candidate 1) (ubuntu lts)
22:04 sm amgarchIn9: was that near your max ram ? I saw no high usage on hub so I think your darcs got OOM-killed
22:06 sm maybe egrep -i 'killed process' /var/log/messages will tell
22:06 sm or /var/log/*
22:07 sm egrep -i 'killed process' /var/log/{kern.log,syslog}* , on this ubuntu
22:08 amgarchIn9 I dont see OOM here. RAM is 8G here.
22:08 amgarchIn9 Pushing one by one works actually
22:10 sm I'm glad I found these.. now I can see when darcshub & ssh server got killed and big they were
22:13 sm hub's ssh server was killed shortly after your push earlier (lock file created 03:42 PST, server killed 03:49) so I think the problem on my end is (again) high memory usage triggering oom killer.. which is a bit indiscrimate.. sometimes it lets the (smaller) darcs subprocesses live (so I see stray processes), sometimes it kills them too (so I see only their lock file). Good to know
22:15 amgarchIn9 sorry for that
22:15 sm thank you, but no need to sorry, it's good testing
22:16 sm are some of those 1800 patches you were trying to push, large, eg involving many files ? darcs changes -s would show
22:17 sm I'm trying to figure out what to tell people to avoid in the FAQ
22:19 sm eg "pushing more than 100 patches, or more than one patch containing more than 100 files, at a time may fail ?"  (that would be lame, but maybe that's how it is ?)
22:21 sm by the way darcs hackers, how hard is it to make darcs push as memory efficient as darcs pull ? (I always hear darcs pull works better in these cases)
22:21 amgarchIn9 du -sk in _darcs/paches shows 8M the largest patch
22:21 sm amgarchIn9: thanks, and how big are all patches together ?
22:22 amgarchIn9 whole directory is 20M
22:22 amgarchIn9 they ar gzipped, I assume
22:22 amgarchIn9 *are
22:22 sm I'll do some testing with numbers like these
22:26 sm and add it to the faq. Thanks amgarchIn9
22:27 amgarchIn9 when I do one by one I get occasionally "Apply failed". A few retries and it works. Does hdn need some time to complete requests?
22:28 sm amgarchIn9: hmm. you're doing darcs push of one patch over ssh right ?
22:28 lelit` joined #darcs
22:28 amgarchIn9 may I start "while true; do echo yd | dvs push hdn:repo; done"?
22:29 amgarchIn9 yes
22:29 sm sure
22:31 sm no high load here, but it seems to be pushing slowly
22:32 sm as if you had a sleep 60; in that loop :)
22:34 amgarchIn9 there is one that consistently gets "Apply failed". I terminated that loop
22:35 sm I don't think I have good enough logging here to say more about that
22:35 sm maybe adding --debug-verbose on your end will help
22:35 amgarchIn9 There is an *.so file in there. Binary. (You know people check in all the random staff when they are new). Are there restrictions on file type (securitiy and such?).
22:35 sm no
22:39 amgarchIn9 what kind of patch is that? http://pastebin.com/xj2nzXk2 Adding and removing a binary in the same patch?
22:39 amgarchIn9 http://pastebin.com/Ufg64a4U this is what darcs cha -sp shows, namely adding an *.so file.
22:41 amgarchIn9 ignore that. That is the next patch that removes *.so
22:41 sm I think --debug-verbose is sometimes better than --debug
22:41 sm oh you didn't type that --debug, never mind
22:42 sm amgarchIn9: is the repo available somewhere else public, so I could try ?
22:43 sm also I'm still wondering why your patches are coming in so slowly - do you have a delay in your loop ?
22:48 amgarchIn9 darcs get http://79.214.5.84:/darcs/pts
22:49 amgarchIn9 it is stopped. I do not get past the patch named "local optimiser added"
22:50 amgarchIn9 I can push newer patches though if I skip that one.
22:55 sm getting that repo
22:56 sm sloowly
22:57 saep joined #darcs
22:58 amgarchIn9 thats my laptop, wlan. upstream only ~100K. But you do not may it out. Actually it stopped.
22:58 sm I'm guessing some patches are highly conflicting
22:59 sm but then, surely I'd see high cpu or memory use from darcs get here, which I don't
23:00 amgarchIn9 the early patches are from a single person. There should not be much conflicts. There are definitely some. I remember I had to rebase some staff because of them.
23:02 sm got it
23:05 amgarchIn9 there were pauses. But then you actually did max out 130 kbps upstream.
23:05 amgarchIn9 Are there filename restrictions like dos upper/lowercase insanity?
23:06 sm I forget the details, modern darcs should warn you about them before committing
23:07 sm when I tried to push this repo, darcs push on the client immediately used up all memory
23:09 sm using darcs 2.9.7 (+ 76 patches). I would try an older one, but I don't seem to have any handy
23:10 sm amgarchIn9: thanks for the info, I have to go but will test more. I opened http://hub.darcs.net/simon/darcsden/issue/51
23:10 amgarchIn9 thank you
23:12 sm bens: looking at http://hub.darcs.net/simon/darcsden/issue/48 also
23:49 bens sm: thanks for following it up :)

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary