***boegel is now known as boegel|afk
<mark_weaver>davexunit: what was the problem? I see only one commit pushed from you recently, and I don't see how it could have fixed that problem. <mark_weaver>if you push a change to mesa, it might be better to push it on a different branch and add a jobset to hydra to build it before pushing it to master <{[]}grant>Racket's Raco complains that it needs 'sslv2-or-v3', only openssl 1.0.2 seems to be available though. <davexunit>mark_weaver: I didn't push a patch to fix mesa, I haven't figured it out yet. <davexunit>I did push a patch that was causing 'make' to fail <davexunit>due to the prior patch by iyzsong, which removed a patch file, but not its reference in Makefile.am <davexunit>our mesa package drops i915 and i965 drivers when building on non-intel platforms, but I fear that it may accidentally be dropping nouveau (and maybe some more?), too <davexunit>depending upon what the default dri drivers are built <mark_weaver>well, we explicitly pass 'nouveau' to the list of dri drivers on non-intell <davexunit>but is that just because, without explicitly saying what drivers to build, it builds *everything*? <mark_weaver>but the problem you're having is presumably on an intel platform, where we don't pass that argument at all <mark_weaver>so I'm not sure why you'd be looking at that as a potential problem <davexunit>mark_weaver: it's just the first place I thought to look. the expected driver .so file just isn't there in our mesa build. <davexunit>so I thought maybe it's something we must explicitly ask to be built? <davexunit>there's a nouveau_vieux_dri.so file, but not nouveau_dri.so <davexunit>and that's the file that libGL complains about not being able to find <davexunit>do you think I'm looking in the wrong place? <mark_weaver>honestly, I have no idea. it would require a closer look <mark_weaver>my time spent getting WebRTC working on our Icecat was not in vain :) <mark_weaver>I have to disable LibreJS to make it work, however. even whitelisting the entire site in LibreJS preferences doesn't work (it's not the first site I've noticed that with) <jxself>That means the JavaScript is either not properl marked as free, or isn't free. I wonder which it is... <mark_weaver>but in any case, LibreJS has a bug with whitelisting that will cause many people to have to disable it entirely, which is too bad. <mark_weaver>at present, this seems to be the only thing with the potential to displace skype <mark_weaver>(for people who want to be able to chat with their family and not just other free software folks) <jxself>I wonder if that's needed, since Hello is "Powered by Telefonica" they say. <mark_weaver>apparently the URL "about:loopconversation" is involved as well. not sure how to whitelist that one <mark_weaver>and the URL "wss://media021-sjc.tokbox.com/rumorwebsocketsv2" (or similar) are also involved somehow <mark_weaver>this is based on logging done by the RequestPolicy extension <mark_weaver>(which I don't use but my friend does and he passed along that info) <mark_weaver>but when I click on LibreJS's icon in the Icecat toolbar, which shows a report of what JS was blocked, it said that nothing was blocked. <mark_weaver>if something else is getting blocked that's not listed there, maybe it's just that the report needs to be made more comprehensive somehow <mark_weaver>but fwiw, when I used NoScript, I never had any trouble getting sites whitelisted <iyzsong>oops, I wrote a type in commit message (4ec486). sorry >.< <mark_weaver>obviously, we should try to be careful, but I've made similar mistakes and so have civodul. life goes on :) <mark_weaver>iyzsong: git allows rewriting of history, but as a policy decision we don't allow it on the master branch <mark_weaver>it would cause problems for others who have based work on our repo <sneek>Welcome back wingo, you have 1 message. <sneek>wingo, civodul says: re gcc, you need to install "gcc-toolchain" instead of "gcc" <wingo>sneek, later tell civodul i tried that, but somehow i missed the message of the env. variables i would need to set <wingo>anyway, on my way. terminal, irc, emacs, web browser -- that's mostly there :) *wingo didn't realize hexchat was the new xchat <wingo>i guess i don't understand the meaning of "the following environment variables may be needed" <wingo>if they are really needed then perhaps there should be some .d directory in the profile that ~/.bash_profile can source <wingo>otherwise if i uninstall a package i could end up with a stale entry in my .bash_profile <wingo>is there something special i need to do to get perl modules to work from a user profile? <wingo>upower's configure balks at a missing perl XML::Parser <wingo>iyzsong, i guess? i didn't see any recommendation of what to set it to <wingo>currently looking to find out what that should be <wingo>the issue might be that perl is from the system install and i am using it as a user <wingo>er, installing packages as a user <wingo>perhaps the PERL5LIB question doesn't arise if i install perl as a user <iyzsong>yes, that's the problem. the search-paths is only for user's profile, will not respect the system one. <wingo>but i installed XML::Parser as a user <wingo>actually which perl resolves to a perl in my user's profile <wingo>so maybe there is no user/system problem then <iyzsong>wingo: If you have perl in ~/.guix-profile, then 'guix package --search-paths' should get PERL5LIB. But the problem occurs if I have perl in system profile, a module in user profile, no hint from --search-paths. <iyzsong>I think it's a bug, will fire it :-) <wingo>indeed i have nothing in guix package --search-paths <wingo>thanks for the clue about that command too :) <wingo>and for the excellent feedback on the patches yesterday! <wingo>i'm still setting up a good dev environment here so i'm a bit slow <wingo>no main on this machine yet :P <wingo>yay, configure for upower runs <wingo>how do i convert a base16 sha256sum to a base32 sum? <wingo>there does not appear to be a base16 operator <wingo>does everyone just download things twice when making a package -- first time to get the checksum, the second to build? <alezost>wingo: when you use "guix download ..." the package source is placed in the store and it will not be downloaded the second time <davexunit>I have an icecast patch ready to go, so I wanted to catch you before you tried to package it. :) <wingo>alezost, ah, i was just using "guix build" <wingo>which downloads, then carps that the checksum doesn't match, and so doesn't keep the download <iyzsong>davexunit: I won't do icecast. it depends on libshout? <davexunit>iyzsong: no, I just wanted to make sure you were on better things. :) <davexunit>I haven't pushed the patch to add icecast yet because I was going to do some major refactoring of libxml2/libxslt stuff. <davexunit>but I think I should just push it (after patch review of course) <iyzsong>davexunit: I'm on the way for a full build of gst-plugins-good. the missing are aalib, libsoup (to my suprised, it use php for testing). <iyzsong>no, I think it's a HTTP library for both client and server <davexunit>I thought BeautifulSoup used it or something <df_>gnu.org appears to be down, did they annoy the chinese government or something? :( <df_>ah well, savannah is ok, guess I can build the docs from source <davexunit>the physical host that a bunch of these VMs live on went down *mark_weaver wonders if he should do something to fix the evaluations in the meantime <mark_weaver>we need a better way to detect these mistakes somehow <mark_weaver>bavier: for now, I pushed a commit that creates a stub 'perl-gd-options-passthrough-and-fontconfig.patch' and adds it to gnu-system.am. <mark_weaver>this should enable hydra evaluations, although that package will fail until that file is replaced with the real patch. <df_>I appear to be intermittently getting this error when attempting to install packages: <df_>guix/scripts/package.scm:578:4: In procedure #<procedure 2c032c0 at guix/scripts/package.scm:627:17 (expr)>: <df_>guix/scripts/package.scm:578:4: string contains #\\nul character: "\\x00150 Here comes the directory listing.\\r" <df_>well, not exactly intermittently <df_>it was happening consistently for emacs, but then after I installed some other stuff it worked <wingo>df_: that would seem to indicate that something is trying to compile a regex with an embedded NUL <df_>would the rest of the stack trace help? will have to get a web browser installed in order to paste it somewhere... *wingo a guix noob but not a guile noob :) <df_>so this is all a bit alien right now <wingo>what version of guix are you running? <df_>I am attempting a guix pull to see if that helps but it seems to be taking a very long time to compile <df_>does pull get the latest release or the bleeding edge? <wingo>regarding versions i am not sure but there were a number of commits in february that could have improved this bug <wingo>so i am going to punt on investigating it <wingo>fingers crossed for the guix pull! <df_>now I at least have emacs and a browser I can probably have a go at debugging it myself if it persists <wingo>that's about the point where i'm at :) <wingo>currently seeing if i can package upower so that i can package gnome-settings-daemon so that my brightness keys work :) <df_>hmm, so will running guix pull as a normal user only update it for my profile? <davexunit>each user may use a different version of the guix client <df_>and if I want to update globally? just run it as root? <davexunit>to update the guix in /run/current-system, that takes a call to 'guix system reconfigure' <df_>after pulling as root? <davexunit>the 'guix pull' for root will be super fast because you've already pulled the same version for another user <df_>still getting the error :( <mark_weaver>I guess it's a problem in 'check-package-freshness', while trying to ftp to ftp.gnu.org <mark_weaver>df_: btw, you probably don't want to actually install 'binutils' directly. instead, we have 'gcc-toolchain' that contains 'gcc', 'binutils', and 'glibc', along with something called 'ld-wrapper' that wraps 'ld' such that rpaths are inserted automatically. <df_>ah... well I don't get the error for that package <df_>so I guess I've worked around the problem for now... <mark_weaver>df_: are you behind some kind of transparent proxy for ftp connections, by chance? <df_>I haven't noticed any problems with it before but I probably don't use ftp a lot <df_>I'll try ftping in manually and see if anything weird happens <mark_weaver>I suppose it's possible that your NAT router is doing something unusual <wingo>woo, fixed a distcheck issue in upower <wingo>i have a distro on my machine and it just works and when i need to fix it it's all in guile <davexunit>now, to convinve people that aren't guile comaintainers of this! <mark_weaver>we seem to have convinced a number of non-guile people :) <mark_weaver>I guess that our little ftp client doesn't cope well with whatever your NAT router is doing <df_>I will probably do a little more investigation of my own first <mark_weaver>df_: well, the problem seems to be that NUL characters are somehow getting in the ftp stream, which our ftp client can't cope with because it uses C-based regex to search for things in that stream. apparently you're the first person to run into this problem, or at least the first to report it. <mark_weaver>anyway, I have to go mostly afk for the rest of the day. <df_>ok, will have a look <wingo>mark_weaver: upower tests want to connect to the system bus. they work when i run them directly but not from within the build. the build has dbus as an input but i guess that's just the libraries; dunno. <wingo>should i just not run the test? <wingo>they look for /var/run/dbus/system_bus_socket <wingo>which exists in my environment <wingo>and in "guix environment upower" <wingo>but i guess not in the build environment <davexunit>the build environment is far more restricted. <davexunit>is there any particular reason why we use python 3.3.5 and not a newer version? <davexunit>I see stable releases for the 3.4.x series on python.org <mark_weaver>I think we normally have to just disable tests that rely on system dbus <wingo>i guess it would be possible in theory to run tests in an environment with a system bus service <wingo>but no need to poke that now <mark_weaver>I have a vague recollection that he might have arranged to start dbus from within the build environment in some package, but it might have been a session dbus. <mark_weaver>but we certainly have several packages that just disable tests that need system dbus <wingo>too bad i spent time making tests work :) <mark_weaver>wingo: btw, I notice that I can't change any preferences in gnome-terminal. i guess that's due to the lack of the setting daemon or something? *mark_weaver wants black-on-white text <wingo>mark_weaver, yes when i ran the file in the libexec dir <wingo>it spewed something about only having a memory backend for gsettings <wingo>and that profile settings wouldn't be persistent <wingo>perhaps due to some bug they're not taken at all <wingo>in any case it's either lack of a dconf service -- i don't know <wingo>or lack of gnome-settings-daemon <mark_weaver>wingo: btw, I updated your gnome-terminal patch to take into account the feedback from iyzsong and based on current master. <wingo>mark_weaver, great, thank you! <wingo>i don't have mail on this machine yet so i'm a bit slow <mark_weaver>but I don't know what copyright line to add for you. what email addr do you want to put there? <wingo>let's put wingo@igalia.com, then i'll read the list via work email <wingo>uf, one more hurdle -- installing udev rules.. <mark_weaver>wingo: see the 'udev-service' in %base-services at the end of gnu/services/base.scm <wingo>i am still getting the package to install correctly <wingo>is setting --localstatedir=/var correct? <wingo>for udev i am following pulseaudio <mark_weaver>we set --localstatedir=/var for some daemon packages <wingo>it tries to mkdir -p /var/lib/upower <mark_weaver>wingo: those make rules need to be inhibited somehow <wingo>what should make /var/lib/upower? <wingo>when it first runs, you mean? <wingo>i guess we can make deco create the dir <mark_weaver>some of our services do it at activation time, but that's actually wrong, because at that point only the root filesystem is mounted <wingo>mark_weaver, yes i found that, thank you <mark_weaver>more work needs to be done there to make it easier to set up, or perhaps we should just make it the default which would avoid the issues for now. <wingo>was talking about that with civodul yesterday and he found a bug with it <wingo>i wonder how best to make a patch with immutable dirs owned by some other user <wingo>previous things have been hacky <df_>any idea why I'd get this: tcpdump: live packet capture not supported on this system ? <df_>missing a kernel module or something? <df_>and am I right in thinking that if guix goes looking for packages at ftp.gnu.org it's because they're not available on hydra? <df_>anyway, it looks like my router is doing something dodgy because I see a zero byte in the stream from any computer on this network but not from elsewhere *davexunit pushes initial 'guix publish' patches, finally <davexunit>a lot of work left to make it actually useful, but it's a start. *wingo attempts to build a system with a upower service <wingo>where does dmd store its stuff? <sneek>Welcome back civodul, you have 1 message. <sneek>civodul, wingo says: i tried that, but somehow i missed the message of the env. variables i would need to set <civodul>so did 'gcc-toolchain' eventually did the trick? <wingo>civodul, yes, though i misunderstood the bit about the env variables <wingo>it would be nice if there were a directory of things to source in the profile <wingo>for sourcing environment variables <wingo>eventually i will go over this log and send you things <civodul>there's 'guix package --search-paths', but i imagine it's not something one immediately finds <wingo>currently i have a building upower package and service, trying to get a system working with a upower system service <wingo>only missing one bit i think, though hampered by lack of logs in qemu <wingo>is there a way to get to tty1 in qemu? <civodul>yes, try C-M-1, i think, to get to the qemu montior <civodul>from there, you can try "sendkey ctrl-alt-f1" <civodul>and then C-M-1 again to get to the actual system screen <civodul>(it's muscle memory though, so i'm not sure about the key bindings) <davexunit>civodul: does our /etc/profile or whatever automagically take care to eval the result of 'guix pacakge --search-paths' ? <wingo>i guess #:initialize was defaulting to #f because qemu images now ask for keyboard input at startup <wingo>gnome-settings-daemon here i come <civodul>davexunit: no, but it should do something like that i guess <wingo>does hydra have a privacy policy? <wingo>it gets hit every time a guix user does anything, almost... <civodul>it's has improved lately, but that's not perfect <civodul>the thing is, when packaging something, it checks on hydra whether it's available there <civodul>but by definition, in that case it cannot be available there <civodul>so the best is to use --no-substitutes here <wingo>given that hydra has to be used at least guix can get the benefit of knowing how many users it has <wingo>kindof an anti-feature for some but it's interesting from the project's perspective <wingo>s/has to be used/is very convenient to use/ <civodul>in the ideal world, hydra.gnu.org will no longer be a central point <civodul>davexunit: BTW, any reason why you didn't push 'guix publish'? :-) <davexunit>if you mean why it took me so long, it's because I was too busy to address your final patch review. *wingo sent alt-kernel and upower patches <davexunit>I packaged a neat little Python program called 'livestreamer' <davexunit>so you can watch live streams from twitch.tv and stuff without using proprietary software <davexunit>serves the same purpose as youtube-dl does for static videos. <wingo>la la la to build the latest gnome-settings-daemon i have to upgrade gtk+ and glib and cairo, whee <{[]}grant>davexunit: Oh yeah, that's a nice bit of software. I've used it to stream from ustream for a podcast or two a few times. :^) *{[]}grant is repurposing the defunct/non-applicable 'buildbox', into just a general Guix(SD) development box. He really wants to get into at least package contributing again relatively soon. <{[]}grant>I have a list of like 10 things I want done, only about 3 started, and 3 that will surely have amazingly long ' *{[]}grant really needs into learning how to write a dmd-service; It amazes me that we don't have an openssh service yet. <wingo>you just look at another service and monkey-type <{[]}grant>wingo: Well I'm in a perpetual state of "monkey-type". <wingo>glib tests take a while to run *civodul is always afraid when it comes to running those large test suites <davexunit>what's the easiest way to test new services? <wingo>davexunit, guix system vm with a minimal os config <wingo>if you don't include x, that's better <wingo>because you can see the dmd output which goes to tty1 <wingo>without mucking about the the qemu monitor <civodul>wingo: re rebuild, we usually use a separate branch and have hydra build that branch <wingo>even for minor glib upgrades? <civodul>probably, because there are 354 packages depending on it per "guix refresh -l glib" <wingo>cairo too is annoying in that regard <wingo>anyway, mailed glib and cairo patches, fwiw <wingo>i'm not yet subscribed so they might be in the queue <civodul>i'm doing terrible profiling in a branch <civodul>the least i can say is that there is no premature optimization :-) <wingo>lol that i am rebuilding my x server <wingo>at least the x server does nothing these days <civodul>davexunit: Icecast is GPLv2-only, right? <wingo>could it be that our xgettext is not actually linked against expat? <wingo>the new gtk+ build is carping about xml support being unavailable <civodul>something must be wrong with our gettext recipe <wingo>i am poking adding a --with-libexpat-prefix <wingo>dunno if it is necessary tho <wingo>oh gods this is at the root of everything isn't it <wingo>it's downloading bootstrap binaries :/ <wingo>on the plus side i'll never need to download anything ever again :) <civodul>gettext is one of the core things, indeed <wingo>does rebuilding it rebuild the bootstrap binaries? <wingo>i think i just rebuilt gettext for nothing <wingo>wishlist: guix environment that could chdir me into the source <wingo>how do i get the source for something that built correctly? <civodul>but i'm not sure that's what you meant <wingo>apparently the default is to attempt to dlopen gettext <wingo>i guess it's to decrease runtime or something... <civodul>i think we need to reinstall the link-expat phase removed by that commit