IRC logs for #baserock for Wednesday, 2015-08-12

*** paulw has joined #baserock06:31
*** rdale has quit IRC07:21
*** rdale has joined #baserock07:21
*** petefoth has quit IRC08:02
*** petefoth has joined #baserock08:02
*** bashrc has joined #baserock08:03
* tiagogomes notes that openssh 7.0 was been released
rjekHandful of security-related changes, but nothing burning08:24
*** edcragg has joined #baserock08:40
*** jonathanmaw has joined #baserock08:41
* nowster goes mad.10:52
richard_mawwhat's up?10:52
nowsterwhat the hell happens here:10:52
nowster  artifacts:10:52
nowster    gcc-libs: build-essential-minimal10:52
nowsterwe're within build-essential, and making gcc10:53
richard_mawthat's saying that the gcc-libs chunk artifact should be put in build-essential-minimal10:53
nowsterWHAT THE ****!10:53
nowsterso what's the products table at the top of build-essential doing?10:54
nowsterand how do they conflict?10:54
richard_mawthe one at the top is pattern matching, so you can say that every chunk artifact ending with -devel goes into a different stratum artifact10:55
richard_mawthen IIRC the entries in the chunks list override anything given above10:56
richard_mawsince it's a more specific assignment10:56
nowsteryes, but why have a reverse mapping in the chunks table?10:56
nowsterit's the wrong way round for a start10:56
richard_mawis it? chunk artifacts can only go into one stratum10:57
nowsterhaving the same thing defined in two places is just wrong10:58
richard_mawif you had it the other way around you'd need to make it a mapping from stratum artifact to list of chunk artifacts10:58
nowsterdo it once10:58
* richard_maw checks definitions for context10:58
nowsterso build-essential defines matching rules and then extra mappings10:58
richard_mawit was a trade-off between verbosity and the need for a more specific assignment11:00
richard_mawI haven't yet gotten to the discussions about reworking the format, but making artifact splitting easier to understand was on the list of things I wanted to do11:00
richard_mawI'm sorry for all unnecessary stress my design has caused you. I tried to get more feedback at the time.11:03
* nowster needs some more coffee.11:04
* richard_maw goes to grab some lunch11:05
*** CTtpollard has quit IRC12:37
*** CTtpollard has joined #baserock12:52
*** petefoth has quit IRC14:19
tiagogomeswhat people think about splitting morph logging output in two files: one with the logging of the last morph command executed, and another with the the log of the previous commands14:43
tiagogomesthe reason is, everytime I need to look at the morph.log, I need to remove it and re-run morph, as the file is too big14:44
richard_mawso one file for all logs, and another for the latest command?14:54
richard_mawjust to clarify, the log of everything will also have the last command in it?14:56
tiagogomesis that relevant? Although if not I'd need to add startup code to copy what is on current to the all-logs file14:59
richard_mawyeah, that was what I was worried about, I'd rather it went into both at the same time14:59
nowsterbug in strata/build-essential/glibc.morph...15:00
nowster- artifact: glibc-bins15:00
nowster  include:15:00
nowster  - (usr/)?s?bin/.*15:00
nowster  - (usr/)?libexec/getconf/.*15:00
nowster  - (usr/)?lib/libSegFault\.so(\.\d+)*$15:00
nowstershould be15:00
nowster- artifact: glibc-bins15:00
nowster  include:15:00
nowster  - (usr/)?s?bin/.*15:00
nowster  - (usr/)?libexec/getconf/.*15:00
nowster  - (usr/)?lib(32|64)?/libSegFault\.so(\.\d+)*$15:00
pedroalvarezthat change makes sense15:01
tiagogomesrichard_maw, that the easiest way to implement it. But it will make a bit more difficult if you want to compare the log of the last command with the log of the second to last command. Not sure how useful that would be though15:03
*** persia_ has quit IRC15:06
*** persia_ has joined #baserock15:09
*** persia_ has joined #baserock15:09
*** paulw has quit IRC15:10
tiagogomesnowster can you send a patch? :)15:11
tiagogomesdid anyone profiled morph?16:05
*** jonathanmaw has quit IRC16:07
paulsherwoodtiagogomes: have you looked at ybd log output? i've tried to make it as helpful as possible16:07
paulsherwoodtiagogomes: what do you mean by profile?16:07
tiagogomespaulsherwood I haven't16:07
paulsherwoodi did build time for morph at various times16:08
tiagogomespaulsherwood run morph in a code profiler and see which lines/functions consumed more CPU16:08
paulsherwoodah, noty.16:08
tiagogomeshi, I am thinking in further improving  morph debugging after the series. This will possibly change code the was modified on . Should I abandon that patch or submit another separately? if it didn't have already a +1 I'drop it16:17
tiagogomesAlso, could be merged16:18
richard_mawtiagogomes: could indeed be merged, though it's currently suffering from path conflicts16:20
tiagogomesok, I'll rebase it16:21
richard_mawI'd prefer if we merged first…16:22
richard_mawtiagogomes: I'll give it another read through, and if I still like it I'll +216:22
tiagogomesrichard_maw probably needs to be merged first before than
tiagogomesotherwise at least the chroot scripts will be broken16:24
richard_mawtiagogomes: reviewing16:26
*** persia_ has quit IRC16:42
*** persia has quit IRC16:42
richard_mawtiagogomes: reviewed: you have a fd leak in
richard_mawthough it would take a huge number of concurrent morph builds to trigger a gc fail because of it16:44
paulsherwoodcan morph actually run concurrently?16:45
*** bashrc has quit IRC16:47
richard_mawit used to be possible, provided morph didn't share a cachedir or tempdir16:49
tiagogomesrichard_maw ok, I'll look into it tomorrow. I may just add a `if fd != None: os.close(fd)` to the except statement16:50
tiagogomesanyway, the current code is an improvement over the other code, as staging areas created when morph crashed will now be cleared with `morph gc`16:51
richard_mawtiagogomes: you'd have to nullify fd first, but if you widen the range of handled exceptions, that should be enough16:52
tiagogomesrichard_maw that was the plan :)16:52
richard_mawtiagogomes: oh yes, it's definitely an improvement16:52
*** tiagogomes has quit IRC17:01
nowsterdoes anyone know why the difference between autoconf-tarball in build-essentials and autoconf in core?17:08
nowsteractually, they're both in core17:09
nowsterah, right... bootstrapping17:09
nowsterignore me! :)17:09
*** persia has joined #baserock17:24
*** persia has quit IRC17:24
*** persia has joined #baserock17:24
* SotK thinks about poking around at storyboard-on-baserock18:45
SotKDid anyone ever get anywhere looking at theat?18:45
SotK(I think persia may have looked at it once upon a time, though I may be wrong)18:45
*** edcragg has quit IRC18:51
* SotK is delighted to discover that apache and mod_wsgi appear to already be in definitions18:53
*** paulw has joined #baserock20:45
*** paulw has quit IRC21:15
SotKSadly, mostly all I've discovered is that its going to need a big reorganisation of the openstack-* strata21:16

Generated by 2.14.0 by Marius Gedminas - find it at!