IRC Logs for #crux-devel Saturday, 2019-05-18

Nomius< stenur> Nomius: i use apulse, as in "apulse firefox"; luckily works fine! <---- hehe, yeah, tried that, didn't work pretty well :-(02:00
*** jue has quit IRC10:34
*** jue has joined #crux-devel11:20
TimB_so, on a serious note: I have been waiting for tek to answer a mail for 7days now, several weeks if we consider the flyspray tickets as well. can somebody help me? I want to move stuff to the 3.5 ports tree, starting with all of my libreoffice ports
*** stenur has joined #crux-devel12:53
ryuoTimB_: sometimes i think the sun will go supernova before tek_ gets around to doing anything. =p12:55
TimB_ryuo: 2020 then? :P12:59
Romsterwhat got less than a year to live12:59
Romsterhe must be smart why am i here.13:00
jueRomster: hmm, is it actually necessary for gobject-introspection to download the glib sources?13:08
jueTimB_: don't panic, it's still a hobby and sooner or later we will hear something from him ;)13:12
Romsterjue, hmm not entirely sure i'll test again without the glib sources13:13
juebtw, add meson and ninja to the deps, please13:15
TimB_jue: not panicing, just a bit bummed about it? the state of opt/texlive is just bad for example.. just why?13:15
jue... and ?13:16
Romsterwiating on rust compile so i can fix rust-bin for 3.5 cairo pulls in meson and ninja i did and findredundantdeps -s -f $(basename $PWD) removed them, though i think i can revert off cairo and back to glib and add ninja and meson. and remove the glib source. going what lfs ues13:16
TimB_and that's bad, why should it stay that way?13:16
Romsteri took libity to fix openjdk8 sources on behalf of j_v until he fixes his site.13:18
jueTimB_: it should not, but over the years I have learned to accept that people are coming and going ...13:18
TimB_I need texlive a lot, so I need it running. I fixed it. it could be better very easily. erase my name from the Pkgfile, slap teks name on it, I don't really care about that aspect. But it seems to be a "nobody touches my stuff" game for that part and I don't understand it13:19
jueand, TBH, I don't use/care about texlive13:19
juewell, you have a working texlive, so what's the problem?13:21
TimB_maybe there are users that would want that but don't look into flyspray or aren't able to fix it for themselves.. I don't see a reason why it should stay that way until tek has time for it.13:22
TimB_but sure, erease texlive. still I would love to get my repos straight, and merging libreoffice would be a great step in that plan13:22
jueTimB_: as I said above, don't panic :)13:24
TimB_well I guess I can go back to sleep then13:26
pedjaRomster, apparently spirv-tools breaks shaderc and vulkan-validation-layers, at least on my system13:28
Romsteryou did read the commit message ML thing about updating spirv-headers before spirv-tools?13:29
pedjanope :)13:29
Romsterat compile time or running time?13:29
pedjaI've been busy painting the house for a couple of days, no computer stuff13:30
Romsterwell i had to add more of the header files for spirv-headers for spirv-tools to be happy13:31
Romstera sysup would might break if it didn't update headers before tools13:31
Romsterpretty sure prt-get sysup resolves the order of updating as it does deptree13:31
pedjanot that bothered, tbh, I don't actually need vulkan for anything yet13:36
jueTimB_: just wrote a email to tek_ asking him about his status, let's see ...13:39
Romsteri'll double check pedja but mpv is working still with shaderc here13:39
TimB_jue: alright, good news. maybe he will react to yours rather than mine13:39
Romstermpv is my main use currently wine not so much yet for dx1213:39
Romsterprt-get update -fr spirv-headers spirv-tools glslang shaderc13:41
pedjaRomster, out of curiosity, why are you building static and shared libs separately in glslang port?13:41
Romsterthat seems to be how it is done. and it's in opt so it should provide both.13:42
ryuopedja: usually that's done because not all ports support building both in the same build configuration.13:42
ryuoerr packages rather13:42
ryuobecause static and shared library object files require incompatible flags to build, afaik.13:43
ryuothough i don't think i've ever tried to use object files from a shared library in a static library.13:43
pedjaryuo, in case of glslang, it builds 2 static and the rest shared with -DBUILD_SHARED=ON13:44
ryuoOh, I see.13:44
Romsteryou got a Pkgfile of your method pedja ?13:46
Romsteror you only do shared?13:46
pedjaI keep it simple :) just a sec, I'll paste it13:47
Romsteri thought i was doing it simple13:47
*** kyaaa has quit IRC13:48
Romsteryeah so you wont have any static libs pedja13:48
*** kyaaa has joined #crux-devel13:48
pedjatbh, I don't see the point in shipping them unless some application requires them13:48
Romster-- Packages updated13:48
Romsterprt-get: updated successfully13:48
pedjaand, for now, none of them did, afaict13:49
Romsteri put them there in case something requires them also opt should ship static versions of the libraries as well.13:49
Romsterisn't that correct jue ?13:49
Romsteri also fixed libcroco and librsvg removed --disable-static13:50
jueRomster: yes13:52
Romsterso what i did with glslang is correct to ship also static libraries13:52
Romsterbecause that's how glslang works for static and dynamic, on different builds. as ryuo said some need to build twice for each.13:53
pedjaI am curious why the upstream default is to build just static libs13:53
Romsteri haven't got a clue pedja13:54
Romsterin boost i can do link=shared,static and get it all in the one build/compile and yet with fftw i have to build with different options for each library13:55
Romsterjust how things go13:55
pedjaI guess different rules for core/opt. I generally only ship static libs if some application complains loudly about them :)14:04
Romstersame pedja to be hoenst i would prefer dynamic for everything. but dmsetup and crypto.static for obvious reasons for initramfs use.14:07
juepedja: currently we have no policy to remove static libs14:07
Romsterjust having the static libs there doesn't hurt anything though.14:07
juewell, sometime it's handy to build a static binary14:07
pedjamakes sense to err on the safe side14:07
Romsterexcept some compile (mostly linking) time14:08
Romstererr okay vulkan failed to build but everything else vulkan-* built.14:08
pedjajue, true. I build veracrypt static.14:09
Romstervulkan is owned by frinnst14:09
pedjais he a Bond villain?14:10
pedjasharks with lasers on their head guarding it, too14:10
Romsteropt/vulkan is validation layers, missing description missing URL old version14:11
RomsterSignature file not found, line 12: cd: vulkan- No such file or directory14:11
Romsteris my ports tree dirty...14:12
Romstermust of been... i removed it manually and ports -u and it's gone so odd.14:14
Romsterok so it all compiles fine mpv is working with vulkan pedja14:14
pedjahm. spirv-tools has 2 pkgconfig files for spirv-tools, spirv-tools and spirv-tools-shared14:21
Romsterfor static and shared linking i would gather14:22
Romster.a and .so files are there.14:22
pedjaone and the rest are static14:25
Romsteri see that, they must have reduced number of funcitons on what you want linked in for static builds.14:26
pedjabut, with build_shared_libs=on, there are no static libs, just the one big and separate tools-{link,opt,etc)14:27
Romsterthough i thoght you only link in functions that you use.14:27
pedjaiirc, shaderc devs asked of them to expose some more functions in a dynamic libs14:28
Romsterso with build_shared_libs=on it disables static files in spirv-tools14:28
Romsteri don't specify build_shared_libs=on in spirv-tools so it builds both.14:29
Romsterkinda odd way of doing so, you would think build_static=off would do that same behaviour.14:29
pedjawhen built as a part of the glslang, same thing, just the .so libs for spirv-tools14:30
Romsterhmm so if i removed -DBUILD_SHARED_LIBS=ON on glslang it'll build static and dynamic in one pass?14:30
Romsternot tested this14:31
Romsterjue, pushed changes to gobject-introspection in 3.4 still gotta do 3.5 waiting on rust-bin atm for 3.514:35
Romsternope pedja that just removes shared libs when i don't set -DBUILD_SHARED_LIBS=ON for glslang14:42
Romsterhas to be done the way i am to make both14:45
pedjaRomster, not even
pedjaI understand you have to jump thru hoops to build both, I am just wondering why the need for both, that's all14:46
Romsteri was just testing glslang14:47
Romsteryou want me to add -DBUILD_SHARED_LIBS=ON to spirv-tools?14:48
pedjathat's your decision, I have my own port to play with :)14:48
Romsteri ma trying to understand what you are getting at.14:49
Romsteri am*14:49
Romster-DBUILD_SHARED_LIBS=ON to spirv-tools removes the static libraries14:50
pedjaI'd just ship dynamic libs by default, and if the need arises, make a separate static-libs port14:50
Romsterhmm but adds usr/lib/libSPIRV-Tools.so14:50
Romsterah i see it compiles static by default, just like glslang does...14:51
Romsterwhy odes this project do this14:51
pedjahard to tell at the moment which libraries are the preferred way14:51
pedjastatic, separate .so's, one big one14:53
Romstervulkan-loader is dynamic and no static14:53
pedjaI was referring to spirv-* mess :)14:53
Romstervulkan-validation-layers has static and dynamic14:54
Romsteryeah since that i was looking at the others too.14:54
Romsteri'll just build spirv-tools twice for static and dynamic14:55
Romsteramazed glslang didn't complain, but then it's not using spirv-tools just yet14:56
pedjasplitting it up to 2 ports makes more sense to me, but do as you please :)14:56
Romsteri dunno ot14:57
Romsterit's the same program14:57
pedjawell, I don't know the reasoning behind 'if shared libs, build only them'14:59
Romsterthey prefer static so it seems and building shared means they don't build static.15:00
pedjawith static as default, it would make more sense to build shared as well when enabled15:00
pedjabut idk15:00
pedjamaybe the goal is to make it more portable with static, or something15:02
Romsterprobably less ABI breakage with static15:02
Romsterbut i can only assume that15:03
pedjadidn't think of that. maybe :)15:03
Romsterno big deal revdep15:03
pedjaGithub issues are full with people trying to package the damn thing and keep their sanity in the process :)15:04
pedja'known-good-hash-commit' snowflakes15:05
pedjatbf, upstream is aware of the issues15:05
Romsteri must admit this wasn't he easiest to decipher15:06
pedjaif you want a true challenge, try rocm :)15:06
Romsteruh do i really want to? no idea what that is. and do you use it?15:06
pedjaamd gpu opencl compute stack15:07
pedjaI have no hw capable of it, so I just read about it15:08
Romsteri see15:09
pedjaif you thought vulkan is a mess, you are in for a treat, my friend15:09
Romsterdo you want python boost ghc and fortran stuff to maintain :)15:10
pedjaalready maintain my boost and python3 ports :)15:10
pedjahad to switch to 3.7.x because of blender15:11
pedjapretty straightforward, tbh15:11
Romsterso what changes? i could integrate them in?15:12
pedjachanges to ports?15:12
pedjanot really, you already fixed that symlinks issue, iirc15:14
pedjain python315:14
Romsterah yes i did add more symlinks into python315:14
pedjaother then that, it's just version bumps, for now15:14
Romsterands the pkg-config file. i hadn't done python3 for boost yet though15:15
pedjaI have py3boost port, if you are interested15:15
Romsteri was thinking of integrating that into boost itself. since python3 is being used for mesa and stuff now.15:16
Romsteryes please i hadn't got to that yet15:16
pedjait's a mess, tbh, but it works, afaict15:18
pedjaluckily, applications depending on it are moving away from it to something else, so I can soon get rid of it, hopefully15:19
pedjaif only boost used some saner build system :)15:21
TimB_I use that port in my libreoffice repo :) but would be great if it were included with boost itself15:22
pedjait's possible there is some nasty, hacky way to do it :) but I never bothered15:24
pedjapython3 updates are more frequent then boost, so rebuilding the whole thing just for that is wasteful (and slow on my rig)15:25
Romsterspirv-tools update pushed for dynamic libs15:26
Romstereveryone want's the insane build systems15:28
*** dlcusa_ has joined #crux-devel15:31
TimB_pedja: good point15:31
Romsteri'll make a python3-boost port15:33
Romsterbut ti's getting late, finish off gobject and rust then i'm gonan sleep15:34
*** dlcusa has quit IRC15:34
TimB_I have to bounce too, got plans this evening15:34
pedjayou have a life outside this? showoff15:36
*** stenur has quit IRC15:42
frinnstRomster: feel free to pull your 3.4 changes into 3.5. I've 'fixed' a lot of your conflicts and im unsure I did it right16:12
*** xor29ah has quit IRC16:12
frinnstI dont mind it - just that its easy to fuck it up16:13
frinnstso dont blame me afterwards :-)16:13
Romsteri haven't been following 3.5 as closely but i have been merging in changes16:14
ryuowhat's slowing down 3.5 now?16:14
ryuoi thought it would have been out by now.16:14
ryuoi see..16:15
Romstergit blame :P16:17
Romsternever used it though16:17
Romsterwhy is git so slow on opt...16:17
Romsterok rust-bin in 3.5 fixed16:20
*** xor29ah has joined #crux-devel16:20
Romsterdid you just do gobject-introspection when i also did it... that explains the .footprint error stenur mentioned16:25
Romsterso my commit i did didn't show up because frinnst did it before me?16:28
Romsterconfusing git output16:28
Romsteryou did...16:29
Romsterfor the record meson and ninja don't need to be listed on gobject-introspection in 3.5 because glib uses those16:34
ryuoperhaps we need a new feature for prt-get to detect redundant deps?16:35
Romsterwell actually mesa3d does that16:35
Romsteri was doing commit sin 3.4 and 3.5 and frinnst just merges stuff and bumped the version for me...16:35
Romsterthen i get confused as to what happend.16:36
Romstermeh fixed now anyways16:37
*** onodera has joined #crux-devel16:41
frinnsti merged 3.4 to 3.5 yes16:44
frinnstbecause I had stuff to merge myself :)16:44
Romsterahhh but gobject-introspection was at 58 not 60 in 3.416:46
Romsterbut i was waiting on rust-bin to finish to then build that before i committed that.16:47
Romsteroh well sorted now anyways16:47
Romsteri need sleeps16:47
frinnstthen it was 1.60 in 3.5 already16:47
Romsteri must of already commited it oh and then you merged in the deps from 3.4 now that makes sense.16:49
frinnstof course16:50
Romsteram kinda tired i shouldn't be commit stuff this late, is almost 3am16:50
Romsterthunderbird is compiling and i need to sleep now16:50
*** stenur has joined #crux-devel18:16
*** stenur has quit IRC22:15

Generated by 2.14.0 by Marius Gedminas - find it at!