IRC Logs for #crux-devel Tuesday, 2019-04-30

jaegerI don't have a strong feeling about pkgconf vs. pkgconfig - just saw that the bug was marked as 'due in 3.5'01:36
jaegerRomster: rust-bin looks to be missing some libs, and, of different versions than those on the system... so it'll fail revdep checks and I would imagine cargo won't work04:05
Worksterjaeger, i ran revdep and i made sure i sysup before building rust-bin06:38
Worksterlooks like there is a way to make cargo static on that i'll look at it after work06:50
Worksterjaeger must be trying to use rust-bin on crux 3.5 i would assume. i can't fault it on crux 3.407:02
*** onodera has quit IRC07:11
frinnstTimB_: do you have a flyspray account?08:36
frinnsttwo accounts? tb and tim87?08:37
TimB_frinnst: I should just have one actually08:46
TimB_tb is the right one. not sure if I have created timb87 sometime? x_x08:47
frinnstsame real name on both :-)08:47
frinnstanyways, you're now added to the contrib flyspray group08:48
TimB_my bad I guess x_x tb is the one I stored in my database08:48
TimB_ok, so my plan is the following: I wanted to go through my ports and make a list of what I would put to contrib, might be good to get some feedback here if they all make sense for contrib, after that I will test everything in a lxc container to make sure I get everything right.08:50
TimB_working with git should be fine, I have just never worked with anybody else than myself within a git repository :)08:52
frinnstgit fetch && git rebase origin/<current checked out branch>08:56
frinnstyou will get an error if you try and push with a repo out of sync. so its hard to fuck up. and if you do its pretty easy to resolve :-)09:00
frinnstshit happens. we've all done it at one point. so no worries09:00
frinnst(me more than once)09:00
frinnstnoticed that the doc is a bit outdated09:01
TimB_hehe :) ok, it should be fine09:02
pedjaRomster, out of curiosity, why is rust-bin shipping everything instead of just rustc, cargo, libs and man pages?10:33
pedjalibrsvg sure takes its sweet time to build now :)10:35
ryuopedja: Because the kitchen sink has that many dependencies.11:21
pedjaryuo, C is your language of choice, right? any thoughts on rust replacing it, which is what people seem to think will happen?11:25
pedjaLinus is calmer these days, but I'd still love to read 'can we rewrite the kernel in rust already?' thread at lkml11:28
pedjaand no, I am not starting it :)11:29
ryuopedja: i don't think it will happen. C is too embedded in many places.11:29
ryuoplus Rust is very slow to compile.11:29
ryuomakes C++ looks fast from what i've been told.11:30
pedjaif librsvg is any indication, it is sloooow :)11:30
ryuopedja: in all seriousness, compile times are a real issue.11:31
ryuothey make development times longer because any changes require you to sit through the process again... and again...11:32
ryuothough i guess it doesn't phase C++ people.11:33
ryuome, i hate slow compiling languages.11:33
ryuoC compiles rather fast, the only major bottleneck is the preprocessor.11:34
ryuoit's why i think languages like Go are a more serious threat. ;)11:35
pedjafew of my developer friends are knee-deep in c#, some big project they work on11:38
ryuoC# isn't C... It's closer to Java.11:39
ryuoBut the way things are going I wouldn't be surprised if .NET ends up supplanting JVM.11:41
ryuoOracle's ownership of Java isn't helping matters.11:42
ryuosomeone said "They couldn't get Dropbox to run."11:48
ryuoand i felt like saying, "Did you try rolling it?"11:48
ryuoyou raise one BS metric, i raise another. :D11:50
*** xor29ah has quit IRC11:51
ryuowell, it's kinda impossible to measure things like this anyway.11:51
pedjathat one is based on Github projects and questions on StackOverflow thingie11:51
ryuoonly useful to answer the question of11:51
ryuo"Is this language of any relevance?"11:51
pedjathat depends on the domain, right?11:52
ryuolargely yes.11:52
ryuoC is pretty much the domain of native libraries and such. It's not likely to be supplanted.11:52
ryuoThe only serious threat in this area is C++.11:52
ryuoIIRC, Haiku is perhaps interesting for its heavy usage of C++.11:53
*** xor29ah has joined #crux-devel11:54
pedjascientific/math/etc libraries are mostly c++, afaik11:54
ryuonot necessarily. most compression libraries are C.11:54
ryuoGraphics? C.11:54
ryuoat least on Linux anyway.11:54
pedjacloser to the metal?11:55
ryuoI suppose so.11:55
ryuoI was thinking of OpenGL and Cairo.11:55
pedjahow about Vulkan :) ?11:55
ryuoNo idea.11:55
ryuoThat's not used for graphics i believe.11:55
pedjait's a supposed to be a successor to opengl11:56
ryuoi suspect if C++ wasn't so complex it would have taken over to a large extent in userspace.11:57
ryuoit's like the C++ standards bodies were like11:57
ryuo"Yo dawg, I heard you liked your UB. So we gave UB to your UB."11:58
ryuo(undefined behavior_11:58
ryuo(undefined behavior)11:58
ryuowhich is basically standard's lingo for "whatever the implementator wants it to mean"11:58
pedjawait, what? so standard defines 'undefined behavior' as 'whatever'? what are they, teen girls? :)12:08
pedjawhat do actual c++ standards actually cover? is it a minimum compilers must implement so they are compliant or?12:13
pedjaor like rfc's? 'must have', 'may have'12:15
pedjaah, interesting distinction between 'unspecified' and 'undefined'13:27
pedjaryuo, cool site, thanks :)13:30
ryuomy joke was C++ has a lot more UB than C does, simply due to more features.13:31
ryuoUB that can't occur in C due to a lack of the underyling feature.13:31
ryuosome examples of UB:13:32
ryuoevaluation order of function arguments13:32
ryuonull pointer dereference13:33
ryuoaccessing unmapped memory regions13:33
ryuodouble free13:33
ryuouse of freed memory13:33
ryuousing ++ or -- on the same variable more than once in the same expression.13:35
ryuothough most UB is related to memory access.13:36
jaegerRomster: sorry, forgot to specify that. yes, on 3.513:47
jaegerand yeah, looks like it fails to update librsvg because cargo doesn't run14:02
jaegercargo: error while loading shared libraries: cannot open shared object file: No such file or directory14:02
pedjaopenssl-1.0.x issue? why am I not surprised :)14:21
*** onodera has joined #crux-devel15:43
*** onodera has quit IRC19:09
*** onodera has joined #crux-devel19:10
*** stenur has joined #crux-devel20:38
stenurToday i've installed dnsmasq.  It turns out that 3.5 s-s-d cannot handle the package!20:50
stenurI wonder what to do about that.20:51
stenurRevert the Debian changes which cause "matching only on non-root pidfile /var/run/ is insecure"?20:53
stenurI think i should have extended the CRUX patch so that it is ok to use PID file only if only user can write it.20:58
stenurThat is, allow PID file if only the user can write it.20:59
stenurThe documented reasoning (--pid-file of start-stop-daemon(8)) is .. well.21:02
stenurWell i mean it is not true at all _if_ s-s-d knows the user.21:02
stenurFor example, dnsmasq uses nobody, the PID file is 0644, so if s-s-d would know that, there would be no problem.21:04
stenur(You need to become nobody in order to replace the file.)21:04
stenurThe "security risk" arises (i would say) if the daemon fails and an attacker places a different PID, which s-s-d then uses.21:05
stenurHowever, only root can write in /var/run.21:06
stenurSo, if that single and lonely "island of security" is sufficient, extending the patch would be ok.21:08
stenurOtherwise rc.d scripts need to become adjusted.21:09
stenurReasoning on "compromised daemon" is just as true for root, right.21:11
stenurI think i'll extend the patch accordingly now and push it.21:12
TimB_on another topic: I would consider those ports to put in contrib, leave me a message if there is anything not fitting. I would go through each of them before pushing anyway21:22
TimB_whenever I get my rstudio port sorted out I'd put push that as well as the pulseeffects stuff21:22
pedjaah, pulseeffects :) can't build gstreamer plugins so I can't play with that21:26
TimB_pedja: why can't you build them? I could provide packages if you like. but it's still funky regarding user settings21:28
pedjalibjpeg-turbo broke them, iirc21:29
pedjawrt consolekit2, I'd submit that as a replacement for ancient consolekit shipped currently, afaik21:31
TimB_would be fine by me21:32
pedjadon;t think there is a bug for it in flyspray, so when you have some time :)21:33
pedjait being in contrib is fine by me, I can retire my ports for it then21:34
TimB_will create an fs issue in a minute21:35
pedjaTimB_, ostree is a dependency for flatpak?21:35
pedjaI am curious about that, only used appimages so far21:36
TimB_pedja: yes21:37
TimB_ostree, bubblewrap and xdg-portal thing are needed for flatpak21:38
TimB_works pretty good here21:38
pedjasnap is the ubuntu one?21:38
TimB_I think21:38
pedjafamous xkcd comic about standards comes to mind21:38
TimB_flatpak was redhat iirc21:39
pedjabubblewrap sounds familiar21:39
TimB_no idea if it's used elsewhere.. :^21:41
pedjathat's like, what it's called, firejail or something? to isolate applications?21:42
TimB_sounds about right21:42
pedjasome cool stuff you got there :)21:42
TimB_thx :)21:43
pedjaargh. the migraine is kicking my sorry ass. time for painkillers and bed21:44
TimB_gn pedja21:46
*** onodera has quit IRC22:07
stenurHey.  I have pushed an update of s-s-d.  It will reject group and world writable, but allow the rest now.22:22
stenur(To the git repo only, say.)22:23
stenurAlso adjusted the manual.22:23
stenurBut i think it would make sense (also upstream) to allow --pidfile with --user=X, e.g., for dnsmasq --user=nobody --pidfile should be fine.22:27
stenurSome rc.d scripts need adjustment for that then, however.22:28
stenurAnd it is ugly: rc.d script needs info that now is in config file only. :(22:28
*** frinnst has joined #crux-devel23:09
*** frinnst has quit IRC23:09
*** frinnst has joined #crux-devel23:09

Generated by 2.14.0 by Marius Gedminas - find it at!