IRC Logs for #crux-devel Saturday, 2019-06-08

ryuojaeger: did you look at my pkgutils patch yet?00:17
ryuohttp://dpaste.com/0S8EK8Y00:18
jaegerjesus... I wonder if this whole BS is due to dbus using --sysconfdir=/etc now00:25
jaegerI'm gonna feel dumb if that's all it is00:25
jaegerryuo: not yet, sorry. Got a bunch of different things on my mind00:26
ryuojaeger: no rush. i just thought i'd try backporting some improvements to pkgutils.00:26
ryuoa micro optimization to footprint generation.00:26
jaegerI saw the conversation about it, seems like a good pursuit00:26
ryuoif this is approved i'll give optimizing how regexs are used a shot.00:27
ryuoright now pkgutils recompiles them for every match attempt in certain modes00:27
ryuoit's a performance killer if you have a lot to match.00:27
jaegermakes sense00:29
ryuoiirc, it's in how the pkgadd rules are processed.00:30
ryuoit recompiles for every file it has to make a decision about.00:30
ryuoi think i can expand the rule struct to take the compiled pattern.00:31
ryuostore it there and just use that.00:31
ryuodeallocate it once done.00:31
ryuoincreases RAM usage a tad, but RAM is very plentiful these days.00:32
ryuoi think a few megabytes can be spared.00:32
jaegeryeah00:32
jaegeryep, it's all because of the sysconfdir change01:03
jaegerI wish dbus/polkit/consolekit errors were more useful01:04
jaegerryuo: poking at the patch a bit now... first thing I notice is that it doesn't generate identical output01:43
ryuojaeger: before or after sort?01:43
ryuojaeger: i noticed that too, but the way pkgmk sorts the footprint produced identical results after sorting old vs new.01:44
jaegerI'm not doing any sorting, just running yours vs. stock01:44
jaeger'pkginfo -f'01:44
ryuoOh. I see.01:44
ryuoYes, the ordering i know is different.01:44
ryuoi sort the resulting cache files to make faster lookups for hard links later.01:44
ryuoif that is important to keep in original order, i can create a separate lookup table instead.01:45
ryuowhat i was concerned about is identical output after pkgmk does its thing since that is the primary way pkgutils it used.01:46
ryuoerr01:46
ryuofootprint mdoe.01:46
ryuomode*01:46
ryuojaeger: that much i was aware of, but the resulting output should only differ in the order the lines appear as.01:47
ryuojaeger: try running them through 'sort -k3'01:47
jaegerI don't know that it really matters, just noticed it01:47
ryuojaeger: ok.01:47
ryuojaeger: that's what pkginfo does before writing the footprint.01:48
ryuoit's different from how i sort them, at least when dealing with empty files. (heh)01:48
ryuoi didn't think the presorted order mattered all that much because the main user of footprint mode is pkgmk, and it resorts and filters it anyway.01:50
jaegeryeah... off the top of my head I'm not able to think of a reason it would cause any problems. Maybe jue or frinnst could think of something I missed01:51
ryuobut i should be able to match it identically if there's concerns about it.01:52
ryuothe original went through them in archive order and used a hash table for the lookups it needed.01:53
*** xor29ah has quit IRC05:22
*** xor29ah has joined #crux-devel05:28
TimB_jaeger: so you were able to fix the problem? great :)07:30
TimB_tek__: another ping, maybe you have some time on this long weekend?07:31
jaegerTimB_: yeah, fortunately14:08
jaegerSo after 3.5 is released I'll update the main MATE repo with the fixes, plus consolekit2, etc.14:08
TimB_jaeger: jue updated opt/consolekit to use the new upstream :)14:26
TimB_it's probably safe to drop it from the repo, or not?14:26
pedjaI had some issues with libvirt and polkit after that update, but I'll have to check if I fscked something up first :)14:37
pedjait's weird, local polkit rules for libvirt access no longer worked, for some reason.14:41
pedjanow, to try to update to rc2 the normal way. wish me luck :)14:44
*** pedja has quit IRC14:44
TimB_good luck :D14:50
jaegerTimB_: maybe so, will test that at some point15:05
*** stenur has joined #crux-devel15:51
juejaeger: polkit too, it's at 0.116 in 3.5, would be nice if you could test the ports with MATE17:03
juejaeger: what shall we do with rc2 now? Do we care about the init issues found by SiFuh?17:05
jueor just release rc2 with the option to use it as the finale version?17:07
jaegerI do think we should fix the init issue if we can. I'm not sure I fully understand it because what he suggested as a fix doesn't fix it for me17:21
jaegerI don't really care if we fix it before or after the release, though17:22
jaegerYeah, I'll test polkit as well17:22
BurnZeZI noticed some weird locking complaint with llvm when rebooting17:22
BurnZeZSomething about file-based locking17:23
jaegerjue: I wonder if we should replace spidermonkey with mozjs6017:23
jaegerI have no idea if one is better than the other17:24
BurnZeZI meant to ask about this a while back17:25
juejaeger: the updated spidermonkey in 3.5 is build from latest firefox ESR sources, version 60.6.1 ;)17:26
BurnZeZBut has anyone tried making alterations to the build system to use private namespaces?17:26
jaegerBurnZeZ: I think the llvm thing is harmless17:26
jaegerNot I17:26
BurnZeZI thought so too, but wasn’t 100% sure17:26
jaegerjue: I'll give it a look17:26
jueI just kept the old port name17:27
BurnZeZIt would be nice to have repeatable builds without elaborate setup17:27
BurnZeZCould probably use a private namespace and then bind in whatever packages are listed as dependencies17:27
BurnZeZTheoretically you could do this per-user for actively installed things, even17:30
jaegerjue: would you mind looking atthe init thing?17:30
BurnZeZWhat’s the init problem?17:30
jaegergoing to runlevel 1 and then back doesn't work17:31
BurnZeZHmm17:31
BurnZeZWhat happens?17:33
juejaeger: yeah, can try, maybe tomorrow but probably first if I'm back in two weeks17:34
juejaeger: what do you think about our release, can you do it togther with frinnst the next days or should we wait until I'm back?17:36
jaegerBurnZeZ: going to runlevel 1 is fine, exiting runlevel 1 to go back to 2 results in no ability to interact other than ctrl-alt-del17:36
BurnZeZHmm17:37
BurnZeZIs /bin/init not giving up control?17:37
jaegerI don't know what the actual issue is, haven't looked into it yet. Been more concerned with other stuff :)17:37
BurnZeZErr, /sbin/init17:37
jaegerjue: sure, I'm fine with it17:37
BurnZeZWho is the init anyway?17:37
juejaeger: sorry, with what? :)17:38
BurnZeZNot even sure where /sbin/init comes from17:39
jaegerjue: releasing 3.5, sorry17:39
jueok, great :)17:39
jaegerIn fact we could even do it today or tomorrow if you want, I think17:40
BurnZeZOh yeah17:40
BurnZeZI managed to freeze i3/X11 with qjackctl17:40
jaegerhrmm... mozjs builds faster than spidermonkey but I've no idea if there's any functional difference. I'll test it on a full MATE install17:40
BurnZeZSomehow stuff gets locked up until qjackctl dies17:40
jaegerdoh17:40
juejaeger: that's hard, I have to leave in a few minutes and tommorow I have a lot of other things to do, sorry17:43
jaegerThat's fine, I'm sure frinnst and I can do it17:43
BurnZeZmumble crashes as well17:44
BurnZeZWorks fine with jack2, but not jack117:44
BurnZeZ/usr/lib/libjack.so.0(jack_port_register+0x2d)[0x7f7636a2d02d]17:45
BurnZeZmumble(+0x213fe3)[0x561da6f0dfe3]17:45
jaegerjue: have a good vacation :)17:45
juethanks a lot :)17:46
BurnZeZAppears to trigger when jack can’t find the thing in /dev/shm17:46
juebbl17:54
*** pedja has joined #crux-devel22:42
pedjawell, that was fun22:42
pedjalibs no longer shipping .la broke more stuff then openssl-1.1. kind of impressive :)23:13
pedjaopt/libunique is broken, 'warnings treated as errors' thingie23:36

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