<osmano807>Hello. After a long time only looking, I decided to give GuixSD a try. <osmano807>But, I'm stuck trying to configure my trackpad <osmano807>In procedure struct_vtable: Wrong type argument in position 1 (expecting struct): #<procedure 594b4a0 at /etc/config.scm:23:2 (state)> <lfam>mekeor's example is a lot more concise than the one I linked to <osmano807_>lfam: I found a commit 'Make service procedures non-monadic', I don't fully understand why, but I'll review it later ***jonsger1 is now known as jonsger
<osmano807807>Icecat force closing constantly. At least now config.scm parses without error, but the options are not applied to xorg <yrk>can I move the store? <yrk>(don't have enough space on the partition for the builds and stuff) ***jonsger1 is now known as jonsger
<cbaines>Today, I've noticed that /bin/sh seems to have vanished from containers created with guix system container, which makes it harder to get a shell in the container with guix container exec ***snape` is now known as snape
***mildred1 is now known as mildred
<divan>Though I have no idea what code to put where to make this happen. <divan>I'm sure it's quite simple for those that know coding or a bit about guile. Anyone have an idea :) <divan>Or where in the manual/docs can I read up dependencies <thomasd>divan: as I understand that conversation, it means modifications to the Guix code are needed to express this dependency <thomasd>as in, currently, Guix doesn't support dependency relationships between mapped devices, and it should be implemented. <divan>thomasd: Thanks. "Currently dependencies among mapped devices cannot be expressed, but that’s easy to fix (by providing a ‘dependencies’ field as in ‘file-system’.)". That part I don't know how to do. <divan>Apparently there is a simple workaround. But I can't figure that out. :) <thomassgn>you just need the dependencies call with your file-systems declaration of root <thomasd>as I understand it, the "simple workaround" is to implement mapped device dependencies yourself, in a similar way as the file system dependencies :-) <thomassgn>thomasd: you don't I have a similar setup as divan, not for my root, but still a raid with file-system deps. <thomasd>np, other thomas :) The even simpler workaround is to wait for someone else to implement it ;-) it will likely happen, but the question is when... <thomassgn>does anyone know a good way of using my own modules? I mean, where to put them and how to let guile know of them? <divan>thomassgn: I've added this to the filesystem list, but it's not working. I'm pretty clueless at this. Still trying to read through the intro to elisp book. (dependencies (list (mapped-device "/dev/mapper/crypt") <thomassgn>divan: mm, Sorry for being so blunt earlier, but I didn't realize the actual problem you were having was with luks encrypted raid -- which it seems is not yet implemented. <jmd>I have a similar problem: How can I ensure that the network is active before mounting a filesystem? <thomasd>thomassgn: the standard way is adding a directory to $GUIX_PACKAGE_PATH <thomasd>or $GUILE_LOAD_PATH if it's about more general Guile modules (other than Guix packages). <divan>thomassgn: No worries. Any chance someone can update the file-system block of code on that email and paste one with the dependencies added :) <thomassgn>jmd: don't know. Could be possible to do a simple ping test or similar in guile. But no idea how to add it as a dep for FS. But if the FS is a networked one it probably should be able to test for network itself? <jmd>thomassgn: It's not a question of testing. Its a question of specifying dependencies. <divan>thomassgn: that looks useful. Trying something. <thomassgn>is there a way of setting environment vars in guix config? I found schemes setenv, but that's just for current pid or so. <jmd>thomassgn: It is current for the environment where it is called. <thomassgn>divan: is your root on 'crypt'? I'm assuming the luks-dev is on the raid-dev; but I don't see you mounting crypt anywhere. Do you get any errors or do you just not find the device? <thomassgn>jmd: mm, but I would like to put some of my env vars inside the guix config so they are part of the OS; as opposed to having them in .bashrc or something. <divan>and luks-dev is on raid-dev. <divan>I can't recall the error now, I'd have to reinstall to get it. <divan>so the adding "dependencies" to "file-system" object (?), excepts a list. I don't know how to express in the list that first, initialise the raid, create md0, then decrypt /dev/md0 and create /dev/mapper/crypt, then use /dev/mapper/crypt as my root. <divan>thomassgn: think that doesn't help, because it doesn't have two mapped devices and doesn't show how to do the dependencies workaround one needs. <jmd>thomassgn: An environment variable, by definition is bounded by the environment in which it is defined. <jmd>What exactly do you mean "part of the OS" ? <thomassgn>jmd: that I would like to add to PATH in the os config so it is set for all users. Just like some part of the os sets PATH to contain /bin/ in other distros. <jmd>So what's wrong with putting it in /etc/skel/.bashrc ? <thomassgn>jmd: good question. forgot about the possibility, and haven't understood the file-like-objects yet. :P <divan>thomassgn: Hmm. It doesn't give an error. Thanks! I'm attempting an install and will see in a while if it works. Though in `guix system shepherd-graph /mnt/etc/config.scm ` I don't see the dependency. But hopefully the install will work. <thomasd>so once again my patches are stuck in limbo :) I should really decide which e-mail address to use, and stick with it. <yrk>how much space should tmp have in order for guix to be able to compile? <jmd>yrk: I would recommend about 12G <yrk>jmd: ok, thanks. I'll have to resize it then <efraim>My aarch64 board has 2G and 1G swap <efraim>Well, then I think the only time I had a problem with a 4GB tmpfs for /tmp was with llvm <fredmanglis>How would one access the current build directory when writing code for (modify-phases ...) <fredmanglis>It doesn't seem to be one of the available keys in the bag <efraim>You can use 'getcwd' as you might 'pwd' from a console <sneek>Core-updates is fontconfig, eudev and cups-minimal FTBFS, tcsh's test hangs <thomasd>Ok I have no idea what's going on here :) <efraim>I pm'ed him with 'sneek core-updates is...' <sneek>I could be wrong, but staging is none of your business <efraim>sneek_: later tell lfam I taught sneek about core-updates for easier work on that branch. 'Sneek what is core-updates' <efraim>sneek_: later tell lfam ask sneek about core-updates <thomasd>efraim: that's actually pretty handy. do you know who wrote/writes sneek, and where the source is? <efraim>thomasd: no idea actually. sneek was already here when I first found guix <efraim>but he's also in #guile, so he's probably written in scheme <thomassgn>how do I test a package definition? I've got a file pkg.scm with my definition in, but I can't see an way of getting guix build to "target" the file, or guix lint for that matter. guix package -f ./pkg.scm returns nothing, but doesn't seem to do anything either. <jmd>thomassgn: Put it in a directory and point GUIX_PACKAGE_PATH to it. <thomassgn>jmd: that made guix package work a bit, now I receive "guix package: warning: failed to load '(fpm2)'" and "ERROR: no code for module (fpm2)". But it did compile the file first... <jmd>I'm not sure that having that tms in the define-module part isn't the problem. <jmd>Or perhaps you need to delete "guix packages" <jmd>At any rate you need to make sure the name of the module and its path agree. <jmd>thomassgn: Also, I suggest that you remove line 46 <mekeor>yeah, so the module named (foo bar baz) should be in $GUIX_PACKAGES_PATH/foo/bar/baz.scm, right? <jmd>mekeor: I think that is correct, yes. <mekeor>also, pkg-config is usually a native-input. <mekeor>also, the description should be longer than the synopsis. <jmd>mekeor: But he knew that after he ran guix lint <jmd>Has the core-updates merge started? <efraim>The core part of core-updates started its evaluation <thomassgn>sneek_: later tell mekeor thanks for input on the package I was trying to define (fpm2) <thomassgn>jmd: thanks for input. I haven't gotten any of the tools except guix package to do anything with the definition yet. Not sure what I'm missing, something outside of the definition or file I assume. I have changed the file according to your input, but still get 'unknown package fpm2' <sneek>lfam, efraim says: ask sneek about core-updates <lfam>sneek: What is core-updates? <sneek>Its been said that Core-updates is fontconfig, eudev and cups-minimal FTBFS, tcsh's test hangs <lfam>Ominous line from that commit message: "Similar fixes for TCP: " <thomassgn>I will have to look more at it later, have a bus to catch. But I'll see what's been talked about when I get back. <lfam>I would also want to try the commits mentioned in that commit message <ng0>I'm currently searching for a new bot and log viewer (well I have a log viewer already) for gnunet.org. As guix is one of the communities relying on this bot, do you have any special wishes other than reasonable secure and maybe searchable log archives (that's why I'm searching for more solutions)? <ng0>to put this into other words, what do you think right now sucks about the current bot? <lfam>ng0: I would love to be able to search the archives. I can't think of any other wishes <lfam>ng0: I'm curious, what do you think could be improved? <ng0>the format and way the things are stored (part of why it's changing now), other than that I know how using a bot could be avoided, but I can't convince people to switch servers etc.. so I'm comparing some of the well-maintained bots/log viewers or 2-in-1 solutions out there <lfam>guix build: warning: ambiguous package specification `linux-libre@4.1' <lfam>guix build: warning: choosing linux-libre-4.10 from /home/leo/work/guix/gnu/packages/linux.scm:238:2 <ng0>limnoria based, among other things <Apteryx>Hello Guix! Is ifconfig provided by "inetutils" ? Reading the help from ifconfig would suggest so. <sneek>Apteryx, you have 1 message. <sneek>Apteryx, alezost says: AFAIK shepherd doesn't monitor processes at all: it just starts a process and let it live or die, whatever it wants. I may be wrong though <Apteryx>sneek: later tell alezost: There is at least a "respawn" field, which, if I remember correctly, causes watching the "running" slot of the service for its PID, and if it's #f it will attempt to start it again. Don't quote me on that though. <thomasd>Apteryx: if you have installed ifconfig, ll `which ifconfig` will show you where it's from <Apteryx>thomasd: So it's from inetutils indeed :) Thanks! <lfam>Wow, decompressing the linux-libre tarballs is extremely slow. Is this a regression in how they are compressed? <thomasd>so this kernel vulnerability is present when the "DCCP" protocol is enabled. Who or what actually uses this protocol (had never heard of it before) so it's enabled by default on desktop linux systems? :) <thomasd>and is configuring a minimal kernel (something which is very easy to do on GuixSD, don't know about other systems) a useful security practice? <lfam>thomasd: Yes, I think it is. You can't be exploited by software you don't have <lfam>I'm working on patching our "stable" kernels now <thomasd>lfam: I would think so, too. But I don't think I've read the recommendation anywhere (not that I'm very knowledgeable about security stuff). <lfam>thomasd: I think that if one was using Linux in a public facing environment in, as they say, "production", you'd compile a custom kernel that met your needs and didn't have extra stuff. <thomasd>Of course, you can probably also create vulnerabilities with a bad kernel config if you don't know what you're doing? <Apteryx>How is the "label" associated to an input in a package definition used? Is this used to bind the package to a symbol (alias?) <lfam>thomasd: Sure, for example you could choose to disable ASLR because you didn't know what it was <Apteryx>Most of the time they are the same: ("libunistring" libunistring), but it could also be: ("glib:bin" ,glib "bin"). <thomasd>Apteryx: it can be used to refer to the package in the build script, with (assoc-ref inputs "label") <lfam>Apteryx: I can't think of any examples right now, but consider an input like ("libuuid" ,util-linux). You could then use it like ((string-append (assoc-ref %build-inputs "libuuid") "/lib"). <lfam>That would help explain the use of a "miscellaneous" package like util-linux <Apteryx>lfam: OK. So it really is an "alias", to help us humans document/understand what's going on. <lfam>AFAIK, but I'm not a Scheme expert :) <thomasd>I'm signing off. Good night and good luck fighting the kernel bugs. <myglc2>I just noticed that when I run emacs it "peggs" on of my CPUs at 100%. This seems like something new. Has anyone else seen this? <efraim>i think I found a bug, when I compile gcc@4.9 or gccgo on aarch64 it only has runpaths glibc/lib and gccgo/lib, and it should also have gccgo/lib64 <rekado>myglc2: do you get this also when running “emacs -Q”? <rekado>Apteryx: ultimately we would like to do without labels and use gexps instead. <myglc2>rekado: Hi rekado. Yes I also get it with -Q. Interestingly, when I start additional copies of emacs it remains 100% of a single CPU. <rekado>Apteryx: there’s a branch for that but it isn’t ready yet <rekado>myglc2: I suppose you could attach to the process with strace and see what it does <paroneayea>cwebber@oolong:~/devel/pubstrate$ guix challenge guile-next <paroneayea> local hash: 0b42jxhn8rdamqa9nmh91hhgw4ra4m9vafyg9b0mhdxnnqwhxpjn <myglc2>rekado: Which was closed w/o resolution, so maybe I will also generate the feedback he emacs developers requested <myglc2>rekado: ... the feedback _the_ emacs ... <mthl>paroneayea: Maybe the reproducibility fixes have been made only in Guile 2.0? <lfam>I only clicked on that article because xmonad was on my mind due to that email ;) <lfam>So now I'm wondering if we can ship xmonad without ghc <lfam>Debian's package seems to require it, however <paroneayea>lfam: but don't users need ghc or etc in order to customize it anyway? <lfam>paroneayea: I figured as much, but the breadcrumbs in those comments make me wonder <myglc2>rekado: strace shows stead stream 'brk(0x3f05a000) = 0x3f05a000' <paroneayea>> But the configuration is also compiled into the binary. And if you really need runtime configuration without the weight of GHC, well, Haskell is well known as an excellent language to write interpreters in... (I think someone already did that for Xmonad, actually.) <paroneayea>sure, you can write great interpreters in it, but are people using that to configure xmonad today? <paroneayea>and do we have a practical route by which users can use that? <myglc2>rekado: Thanks. It's Oooh Duh operator error, my mistake ;-) <sneek>alezost, you have 1 message. <sneek>alezost, Apteryx says: There is at least a "respawn" field, which, if I remember correctly, causes watching the "running" slot of the service for its PID, and if it's #f it will attempt to start it again. Don't quote me on that though. <alezost>Apteryx: I have no idea what the contents for that is ^^^^ :-) <lfam>catonano: The (gnu packages qt) module <alezost>Apteryx: btw 'ifconfig' can be provided by 'inetutils' or 'net-tools' packages <ng0>I hope no one has pushed an curl update so far <lfam>ng0: I built the update locally and pushed it yesterday <jmd>Do we have a convention for the Texinfo markup to use when refering to other packages in descriptions? <ng0>lfam: I'm waiting with gnURL until the patch release question is solved <lfam>ng0: My guess is that the issue does not manifest with OpenSSL 1.0.2, which is what we are using ***snape` is now known as snape
<ng0>yeah but this is outside of Guix <ng0>I have to think for all systems.. and some might be affected so I rather wait and spare myself the work of doing two releases in short time <lfam>Also, our package doesn't enable SSPI <jmd>If package A can build and run without package B, but really is useless without it, what should be the dependency releationship between the two? <ng0>i think it's easier to say when the situation is known.. <ng0>could be propagated, could be mentioned in the description,.. <jmd>So there is no prefered solution? <efraim>i checked the gcc build code, gcc.scm:86 defines libdir as either "lib" or "out", which means half the files are in lib and the other half in lib64 :/ <jmd>Well I'll post it as is for comment and see who says what. <snape>should I remove the copyright line if I remove some code from some file, and there is no existing code from me anymore? <adfeno>snape: I'm not a lawyer, but my guess is: Yes. <adfeno>Please consult with other more experienced people first. <efraim>when we move packages we also move the copyright notices <jmd>snape: I would have also thought the answer was "yes". <roelj_>Has anyone had the experience of booting into GuixSD, but then get stuck on a grub rescue shell? <paroneayea>is anyone using any cpu frequency scaling tools? <lfam>roelj_: Yes, I've seen that when I configured the file-systems and grub-configuration incorrectly <lfam>paroneayea: No, because we don't have a service, yet ;) <paroneayea>lfam: I had my laptop crash twice last night while compileing guile-next while docked and almost did again while running some unit tests this morning :) <paroneayea>probably should install some sort of cpu frequency scaling thing :) <lfam>I'm curious, how did it "almost" crash? <lfam>Wow! IMO, the CPU should protect itself, but who knows? <paroneayea>I grabbed a fan and stuck it in front of my laptop and crossed my fingers <roelj_>lfam: I think mine is due to installing EFI with a little help from Fedora.. <lfam>I've used CPU scaling before by echo-ing into /sys from /etc/rc.local <roelj_>lfam: Any idea what commands I could use to debug from there? <lfam>roelj_: I hope that somebody else has some ideas. I don't have much experience with the GRUB shell :( <lfam>roelj_: Did you see the previous discussion on the mailing lists about installing on EFI systems? <jmd>lfam: So his CPU suffered a "near miss"! <lfam>Heh, does that mean that is almost missed? ;) <roelj_>lfam: Yes, and I encountered each error exactly in the order it was posted.. :) Well, thanks for your help! <lfam>I guess we still have some work to do for EFI support <jmd>paroneayea: I think, the only thing harder than typing with crossed fingers is driving with crossed legs. <lfam>I am *not* going to try that <taylan>ACTION sees the "leaving the guix project" thread still going stong, shudders :) <jmd>taylan: Yeah. Everyone is saying: "I agree this thread should stop, but I want to be the one who has the last word." <taylan>meanwhile a wikipedia editor has been more or less bullying me because I insisted on defining GuixSD as a "distribution of the GNU operating system." endless drama over a single sentence. <lfam>Man, whatever happened to wikipedia?! <jmd>How is the GuixSD Hurd port going? <taylan>lfam: the more ideological the topic, the crazier it gets, but very generally I think Wikipedia is doing fine <lfam>taylan: I suppose that I only hear about edits that get reverted. Maybe there are lots of edits that are accepted without complaint. I do know that the times I tried to correct typos and grammatical issues in the past few years, the edits were always reverted by a bot. So I stopped contributing. <lfam>Most people I talk to who are younger than 25 years old don't realize that wikipedia is / was open to all contributors <taylan>the coverage of "Linux" is particularly catastrophic though. imagine if the article "Jeep" was reserved for off-road vehicles, and "Jeep brand" used for the brand, "because WP:COMMONNAME" <jmd>Well I used to own a Land Rover and some people insisted on calling it a Jeep. <taylan>right, so Wikipedia should call it that too, as per WP:COMMONNAME! :-) <lfam>I guess it was bound to happen once it killed all the other encyclopedias and became a source of power <lfam>taylan: Please don't drive yourself crazy over this :) ***copumpkin is now known as ikwilclubmate
***ikwilclubmate is now known as copumpkin
<catonano>why don't I see the last thomasd patches in emacs-debbugs ? <snape>I have patchset of 4 patches. I modify one of them (the first). <snape>should I send only this one to the ml again, or all of them? <snape>the other three should apply without conflict <amz3>did some people had a look at janneke talk? what do you think <roelj_>So.. anyone experience with booting GuixSD with btrfs? The error I get is: fsck.btrfs: No such file or directory. <lfam>roelj_: I haven't yet, but I'll try it in QEMU. I want that to work <lfam>roelj_: Yes, and I expected that to make it work. Are you using that code yet? <roelj_>lfam: Was it in the 0.12.0 release? <roelj_>So I probably have to 'guix pull' before 'guix system init /mnt/etc/config.scm /mnt'? <lfam>Definitely, if you are using 0.12.0 <roelj_>Yes.. Would that also solve my problem? :D <lfam>Maybe try it with `guix system vm-image` first, just to be sure <roelj_>Hehe, well I think it'll have to compile a lot so that will take the night <roelj_>Hopefully coming back with a working GuixSD + EFI + btrfs <ng0>can GFDL and GPL3+ be dual licensed? Do manuals/documentations have to be dual licensed if the software is GPL3+ dual licensed? <ng0>I'm trying to find a dual license header for GFDL <ng0>I'Ll take a look, thanks <ng0>i have that offline here, so no need to link :) <ng0>but that's LGPL and GFDL, isn't it? <ng0>I've seen the fixme email with https issue <ng0>I'm not a lawyer, I feel uneasy about adding this GPL3+ + GFDL1.3+ header. But I think revision control, people can fix it <ng0>lfam: thanks for working on it, I had no time to test it <lfam>ng0: No worries, I think it should work if the environment variable works on other platforms. I did test that the variable is set correctly <yrk>ng0: if you are the copyright holder, you can give people a choice under which license they can redistribute your work <yrk>ng0: you can say: you can distribute this work under GPLv3+, GFDL, something else ...\\ <yrk>ng0: this is typically done when documentation contains non-trivial code, so to allow people to use the code in the documentation under a license suitable for software <yrk>perhaps the next version of the GFDL should have a clause which permits further distribution of the work under the GPL <ng0>this is for GNUnet, the new documentation. I expect before this goes live someone WILL double check <yrk>ng0: finally, if you feel like you need more detailed answers, or have an edge-case on your hands, write in to licensing@fsf.org (I'm one of the people who answers questions at that address) <yrk>ng0: thank _you_ for working on free software <ng0>I think if there are questions, Christian or some of the others will get in touch