IRC Logs for #crux-devel Friday, 2019-03-08

ryuojaeger: right. i tried it. you can only open directories O_RDONLY, which doesn't fly for posix style write lock. it'll claim bad file descriptor if you try it. flock though doesn't care. it's a separate construct entirely.00:35
ryuoi'm thinking just to keep the directory style lock with flock.00:36
ryuopkgutils isn't meant to be portable so what does it matter?00:36
ryuojaeger: i also noticed pkgutils starts with opendir, when all it really needs is to store a copy of the FD with the lock...00:38
ryuoso, i'm just going to open the directory myself and use flock on that.00:38
ryuowhy store more than we need?00:38
dlcusa@ryuo, are you aware flocking is merely advisory?00:54
jaegerdon't look at me, I didn't write it :P01:00
ryuodlcusa: yes. for all of them.01:03
*** xor29ah has joined #crux-devel01:45
*** xor29ah has quit IRC01:52
*** xor29ah has joined #crux-devel01:54
ryuofrinnst, jaeger: what do you think I should license this as? I'm open to GPL2 or GPL3 if that's what you think is best for CRUX.07:21
frinnstall crux stuff that is released by us is GPLv307:34
frinnstcheck out ports/(core,opt,xorg,whatever)/COPYING07:35
frinnstbut it's obviously up to you :-)07:35
frinnstare you doing a rewrite? I havent really been following IRC the last week07:37
ryuofrinnst: essentially, yes, but it's in many ways a derivative of the original.07:41
ryuoi'm retaining behavior but rewriting it to be library oriented. it should open up some interesting possibilities for future development.07:41
ryuointegration with scripting languages for one.07:42
frinnstcool cool07:44
ryuoi'm considering revising how the basic locking mechanic works...07:46
ryuowith pkgadd as it is, you have to create the directory structure to even use it on a new root.07:46
ryuoit would be nice if the library could do this automatically when opened in write mode.07:46
ryuoto do that i'd need to lock the root directory as well.07:47
ryuoalternatively, can just keep the existing behavior and require manual intervention to set it up.07:47
ryuoi just recall how pacman never required that...07:47
ryuohm.07:48
ryuoi spent years fiddling with pacman for frugalware.07:48
ryuothis library won't be as complex as that.07:48
ryuothings like dependencies are a foreign concept to pkgutils.07:48
ryuoprobably best it stays that way.07:48
ryuoeh, why not. i'll need to figure out how i'd support it though.07:49
ryuotime to do some charting.07:49
ryuohm..07:50
*** onodera has joined #crux-devel10:28
*** onodera has quit IRC10:28
pedjaryuo, would you consider adding something like 'keep working directory on build failure' as an pkgmk.conf option?11:05
ryuopedja: maybe? i was planning to leave pkgmk largely as is. it doesn't need to be made into a library.11:05
pedja'-kw' does that, but it would be nice to be able to set it globally11:08
ryuoOh. I see.11:08
pedja'-kw' keeps the working dir regardless if the build fails or not, iirc11:12
pedjawhich is a bit of the hammer approach, imho :)11:13
pedjahan's pkgutils fork used that to basically add an option to restart the build from some point, iirc11:15
pedjafor instance, configure/build stages worked, but the cleanup/install failed, because of the typo or something11:17
pedjaso you could fix that, jump over the parts that worked, and restart build from the stage that failed.11:18
pedjacomes in handy with the packages that take a while to build, like qt or similar11:19
*** jue has joined #crux-devel11:35
juehello11:39
ryuojue: welcome back. time to start putting out fires.11:40
ryuo=p11:40
*** xor29ah has quit IRC11:41
juethx, did I miss something important the last week? :)11:43
ryuonot too important. i started working on a library remake of pkgutils.11:44
ryuowith any luck, maybe can be included in 3.611:44
*** xor29ah has joined #crux-devel11:44
jue.oO that's great11:44
*** xor29ah has quit IRC11:51
*** xor29ah has joined #crux-devel11:52
ryuohm. might as well get this whole error system done first.15:09
ryuoi'm finding myself struggling to think of meaningful ones to return.15:09
ryuoguess i'll look at some of the common system ones and start with that.15:09
juebtw, any issues with the new prt-get release?17:47
ryuojue: yes. jaeger ran into an issue with some old patches that changed behavior of prt-get quickdep or w/e when it's not connected to a tty.17:47
jaegersaw one that borked some ISO work, I plan to revert it17:48
jaegerI think it's a bad change anyway17:48
jueahh, didn't see that17:48
jaegerit's a small issue. otherwise no problems17:49
juethe FS#593 commit?17:50
jaegerthat's the one, yeah17:50
jaegerI don't think changing the program's behavior like that is a good idea, personally17:50
jaegerif you look back in the logs we discussed it a bit. given how old it is and how long it sat committed but not released, I think it's safe to revert17:51
jueagreed, not a good idea at all17:52
ryuoi'm considering some improvements to pkgutils behavior on a few fronts.17:53
ryuoif possible, i'm considering making pkgadd/pkgrm support batch operations.17:53
ryuoright now they only support one pkg operation per run.17:54
ryuoanother idea i had for a minor improvement is making pkgadd able to create an empty database in a new root all by itself.17:55
ryuoas it stands you need to precreate the basic structure.17:55
ryuohm.17:58
pedjaerr. shouldn't .nostrip in port dir disable strip? if it should, why it doesn't do that :) ?18:08
ryuopedja: it's just respecting the privacy of sensitive ports. ;)18:09
pedjaheh18:09
pedjait's weird. even thou there is .nostrip, there are a bunch of strip complaining about something messages in the build output18:12
juepedja: you have a regexp in .nostrip? see core/glibc for an example18:12
pedjajue, ah. I thought that just adding an empty .nostrip will disable it. my mistake :)18:14
jaegerI'll revert the prt-get change and make a new tarball today if I can18:18
jue+118:19
jaegerjue: https://crux.nu/gitweb/?p=tools/prt-get.git;a=commitdiff;h=69c857a896d5973abbe95eababe5e6a62097719018:31
jaegerso for a new release I should update the version and make a v5.19.3 tag, right?18:31
jaeger(update the version in configure and configure.in, I mean)18:33
jueyep, a 'make dist-xz' will create the release tarball then19:16
jaegercool, thanks19:16
jaegerlooks like some signature updates are needed for recent commits22:10
jaegerat least for python and vim22:10
jaegerjue, frinnst: I see prt-get 5.19.2 was only pushed to 3.5, was that intentional?22:32
jaegerIf not, I'll push 5.19.3 to both 3.4 and 3.522:32

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