IRC logs for #baserock for Thursday, 2015-07-16

*** drnic has quit IRC03:45
*** drnic has joined #baserock04:21
*** zoli__ has joined #baserock06:30
*** rdale has joined #baserock06:57
*** mariaderidder has joined #baserock07:06
*** paulw has joined #baserock07:13
*** gary_perkins has joined #baserock07:54
*** mdunford has joined #baserock08:05
*** zoli__ has quit IRC08:06
*** bashrc_ has joined #baserock08:07
*** straycat has joined #baserock08:24
*** straycat has left #baserock08:27
*** rdale has quit IRC08:36
*** straycat has joined #baserock08:37
*** ssam2 has joined #baserock08:42
*** ChanServ sets mode: +v ssam208:42
straycathello, i've finished adapting definitions and morphlib/buildsystems.py to use a more vanilla layout for the perl installation. how would we prefer to land it? currently the definitions change i would propose would require everyone to switch to the latest morph08:49
persiastraycat: Is your proposed new layout available somewhere?08:51
persiaOr the related definitions change?08:51
*** zoli__ has joined #baserock08:52
SotKI think if the definitions change needs people to use latest morph, we'll need a release which contains that latest morph doing before the definitions change is merged, and a version bump in definitions.08:53
persiaPractically, yes, but I want to review the nature of the definitions change before having an opinion on whether we ought do that.08:53
persiaBecause if there is a way to can communicate that with existing semantics, we have an easier life: incompatible changes irritate users.08:54
straycatI'm proposing switching to the default (see https://metacpan.org/pod/distribution/perl/INSTALL#Installation-Directories ) using definitions change  https://bitbucket.org/richardipsum/definitions/commits/97e5b243a46cf8cb39864ecf27545a5b5c6f5051?at=ripsum/upgrade-perl-5.2209:01
straycatpossibly losing the definition of vendorprefix, I'm not sure we need that09:02
ssam2fyi, I'm going to propose a way of doing automated migrations of definitions on baserock-dev very soon, and next week or the week after will look at moving default buildsystem commands into definitions.git09:03
ssam2I don't know how any of that will pan out yet, so can't really give useful advice at this point09:04
ssam2but maybe it's useful to know09:04
ssam2the simplified perl .morph file looks handy09:05
straycathttp://sprunge.us/CSUM is work in progress, but as you can see we drop the overriding of all the install paths because we're now using the perl defaults09:06
ssam2looks simpler. But I would have thought that setting PREFIX to /foo.inst/usr instead of /usr would break things09:07
persiastraycat: I like those changes, but I don't understand why this requires updated morph09:08
persiaIs it just the changes from INSTALL_BASE to PREFIX?09:09
straycatssam2, what ExtUtils::MakeMaker means by PREFIX isn't quite the same, http://search.cpan.org/~bingos/ExtUtils-MakeMaker-7.04/lib/ExtUtils/MakeMaker.pm#PREFIX_and_LIB_attribute might help09:12
ssam2straycat: ok, that's good to know. Might be worth adding a comment for the benefit of other people who are used to the 'autotools' meaning of PREFIX09:13
*** lachlanmackenzie has joined #baserock09:14
persiaIndeed: I was also misled09:14
straycatpersia, Sorry that patch is not based off master so may be a little confusing, it's based off https://gerrit.baserock.org/#/c/966/ using --install_base was my mistake, using INSTALL_BASE causes modules to be installed into a different set of paths: http://search.cpan.org/~bingos/ExtUtils-MakeMaker-7.04/lib/ExtUtils/MakeMaker.pm#INSTALL_BASE09:15
persiaHeh.  I think it is worth stealing the comment "INSTALL_BASE can be passed into Makefile.PL to change where your module will be installed. INSTALL_BASE is more like what everyone else calls "prefix" than PREFIX is." from the docs.09:16
straycatessentially morph is doing some work right now to make sure everything goes into /usr/lib/perl, we could stick with that if we wanted and I could modify the Module::Build support I added to buildsystems.py to do something similar, but I just thought it might make more sense for us to have a more standard configuration09:18
ssam2I think it makes sense to have a more standard configuration. I'm struggling to work out the implications though09:20
persiaI mostly wish that there was a sane way to do this without modifying both morph and definitions simultaneously.09:21
KinnisonThere is09:21
KinnisonMove the build systems into definitions first09:21
Kinnisonin a back/forward compatible way09:21
straycatssam2, Instead of all perl modules (both modules included in the perl distribution and modules Baserock installs) being stored in /usr/lib/perl, modules included in the perl distribution will be in /usr/lib/perl5/$perl_version and modules installed by baserock will be in /usr/lib/perl5/site_perl/$perl_version09:22
ssam2straycat: cool, that part makes sense. It's the version of morphs vs. versions of definitions bit where I'm a bit lost09:23
ssam2does it mean that Perl chunks built with "new" versions of Morph won't work with our current build of Perl ?09:23
straycatyes09:24
ssam2right09:24
ssam2so at present, it would have to be wrapped up in a change to the definitions format version. which does indeed highlight that the default build-commands would be much better placed in definitions.git09:24
ssam2which still requires a change to the definitions format version, but a more widely useful one09:24
* straycat nods09:25
ssam2and one I need to do anyway to make it so a useful data model can be generated from definitions.git alone09:25
persiaThen let us do it that way.09:25
straycatActually I'm wrong, the problem I had yesterday was because I was incorrectly using install_base, if I use prefix then the build tools should still use the layout set in perl's configure stage: so if a new morph went to build using the current perl the modules would just end up in /usr/share/perl which would be on @INC so it *should* be fine, even if a bit weird, sorry I was being stupid.09:39
straycatbut if I'm right then this should be simpler, I'll make the changes to morph and verify that we can still build all the perl modules okay.09:40
straycatso I guess morph is just overriding to avoid modules going into /usr/share? i'm still not really sure09:42
*** nowster has quit IRC09:44
*** nowster has joined #baserock09:46
persiaI'm interested, but I really don't understand either.  My suggestion is to test.09:51
persiae.g. try the half of it, and see what happens.09:51
* straycat nods09:57
*** nowster has quit IRC11:15
*** nowster has joined #baserock11:17
*** paulw has quit IRC11:18
*** paulw has joined #baserock11:22
*** nowster has quit IRC11:42
ssam2straycat: I used your virtualenv+pip Python import branch for something and it worked really well11:51
ssam2straycat: i'm curious if you remember any PyPI packages which are good testcases for the Python import tool?11:51
ssam2sphinx seems to have lots of deps, that's a good one11:53
*** nowster has joined #baserock11:57
ssam2oh, TODO.python lists some packages that failed with the old 'pip --list-dependencies' method11:58
ssam2which presumably will all work with the new method!11:58
straycati don't think i've tried any that aren't documented in TODO.python, hopefully more of those will work with the new method, where they fail hopefully it's because we don't search for distributions that are released as zips (which we can change now that we can lorry zips)12:30
ssam2I got 'ImportError: No module named writeexts' trying to deploy something with Morph13:30
ssam2master of Morph, in fact13:31
SotKo.O13:33
*** mariaderidder has quit IRC13:33
*** mariaderidder has joined #baserock13:33
Kinnison:-(13:34
SotKwhat ref of definitions?13:34
straycatssam2, are you using extensions/<extname> in your cluster?13:36
ssam2based off master of definitions13:37
ssam2hmm, I had 'type: tar' not 'type: extensions/tar'13:37
straycat*nod* i had the same problem the other day, it's a little annoying that we have to do this now13:38
*** mariaderidder has quit IRC13:38
*** mariaderidder has joined #baserock13:39
*** paulw has quit IRC13:39
straycatthe error message might be less confusing if we removed the exts from morph13:40
*** paulw has joined #baserock13:43
*** mariaderidder has quit IRC13:57
ssam2straycat: yeah, we could do that neatly if we bumped the version of the definitions format to 6, and then changed Morph so new versions refused build anything older than definitions version 613:58
*** mariaderidder has joined #baserock14:19
*** mariaderidder has quit IRC14:27
*** mariaderidder has joined #baserock14:27
*** mariaderidder has quit IRC14:36
*** edcragg has joined #baserock14:53
*** mariaderidder has joined #baserock15:14
*** mariaderidder has quit IRC15:20
*** zoli__ has quit IRC15:38
*** ssam2 has quit IRC15:49
*** ssam2 has joined #baserock15:50
*** ChanServ sets mode: +v ssam215:50
*** paulw has quit IRC16:00
*** paulw has joined #baserock16:00
*** paulw has quit IRC16:14
*** zoli__ has joined #baserock16:34
*** zoli__ has quit IRC16:36
*** mdunford has quit IRC16:43
*** mdunford has joined #baserock16:57
*** nowster has quit IRC17:00
*** bashrc_ has quit IRC17:04
*** mdunford has quit IRC17:04
*** tiagogomes has quit IRC17:16
*** nowster has joined #baserock17:26
*** ssam2 has quit IRC17:31
*** gary_perkins has quit IRC17:33
*** zoli__ has joined #baserock17:34
*** edcragg has quit IRC17:41
*** paulw has joined #baserock21:37
*** paulw has quit IRC21:44
*** zoli__ has joined #baserock22:41
*** zoli__ has quit IRC22:45
*** juergbi has quit IRC23:21
*** juergbi has joined #baserock23:22

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!