***Guest63232 is now known as sturm
***reepca` is now known as reepca
<atw>sneek: later tell ng0 I think the source uri on colors is wrong <davexunit>well this HN subthread got a lot of attention today <davexunit>including Steve Klabnik from Rust chiming in <davexunit>unfortunately he is one of those "unless it works on Windows it's not a solution" people <davexunit>if you have to use Windows then use whatever you can, I guess. <davexunit>ah screw it. I give up on this. I just can't make binaries that work on other distros <davexunit>holy smokes. I tried one last thing and it actually worked! <davexunit>I got a guix-built guile to load my guile-sdl2 library that dynamically loads libsdl... on ubuntu! <davexunit>wait, nevermind again. it does work. phew... <ovidnis`>the spirit of the computer is just keeping you on your toes :-) ***jonsger1 is now known as jonsger
<efraim>I talked with the guys at #lumina-desktop to clarify the license of the artwork dir, once hydra builds linux-libre i'll try it out in a VM <ng0>Is there a command which could help me find which branch I commited a file to when I can't find it and can't remember the name of the branch? <sneek>ng0, atw says: I think the source uri on colors is wrong <ng0>sneek: later tell atw: possible because I use a shorter source uri format. <ng0>oh, found the branch and commit <efraim>hmm, i'm running into problems with substitute-keyword-arguments and ((#:tests? _) #f) not working correctly <civodul>efraim: perhaps #:tests? is not present in the original argument list? <civodul>in that case, use: (substitute-keyword-arguments args ((#:tests? _ #f) #f)) <efraim>in the past i've used it to enable tests, and I saw it's assumed #t <civodul>notice the third thing in the pattern, which is the default value ***Piece_Maker is now known as Acou_Bass
<efraim>i'll try that, my next try was going to be inspired from hdf5-parallel-openmpi and overwriting the inherited (and declared) hdf5 configure-flags <ransom>hi, all.. i'm a debian adherent beguiled by this distro <efraim>civodul: that worked, qtxmlpatterns successfully built while skipping the tests <rekado>I wished we could simplify overriding the arguments field with “substitute-keyword-arguments” <dvc>ransom: Hi! awesome... :) <rekado>the way it is now we need to know what fields are available in the package we inherit from <dvc>does make pdf work for anyone? make html works, I'm not a texi expert... <rekado>it would be neat if we didn’t have to care about this, resulting in matching exitsing fields or overriding them where appropriate and <ransom>hi, dvc! imma lurk for a moment.. (lurks) <rekado>(inside of “guix environment guix”) <ng0> 0:20.42 configure:15163:23: error: #error "libjpeg-turbo JCS_EXTENSIONS required" is there a problem with this extension that we don't have it? <dvc>rekado: does that mean you have texlive in your profile? Is it still current? Just asking... <rekado>oh, yes, I do have texlive in my profile <rekado>I haven’t touched it in a while, though, but I think there hasn’t been any update to texlive. <efraim>ng0: maybe libjpeg-turbo was built with just nasm to keep the size down? <dvc>if make html works is it ok to commit? <dvc>make[1]: Entering directory '/home/dvc/guix' <dvc> TEXI2PDF doc/guix.pdf <dvc>/gnu/store/mzj6zw3dzil8c0vij2xkwyb3z69mmgqm-profile/bin/texi2dvi: pdftex exited with bad status, quitting. <dvc>make[1]: *** [Makefile:3455: doc/guix.pdf] Error 1 <dvc>make[1]: Leaving directory '/home/dvc/guix' <dvc>make: *** [Makefile:3995: pdf-recursive] Error 1 <ng0>ok, I see if I can find a comment or simply build the jcs extension before palemoon <rekado>have you tried running the command manually? There should be more output. <efraim>I like that the debian-mailinglists include in the header a url to the post <efraim>the oss-security one should do that <dvc>yes, it seems to produce a pdf when run manually <dvc>that is when I run pdftex guix.texi from the doc directory so that relative paths are resolved <ng0>seems like with the new palemoon I'm finally getting somewhere... so what's the problem with APNG for libpng? Simply size, or license issues? <ng0>I see no obvious comments in the file <ng0>I'll ad that to the palemoon graph <dvc>so to solve my commit problem, after the system tests finish running if I still get the same error with HEAD~ I'll push... <efraim>ng0: I've been thinking about your inox patch on the mailinglist, I'm thinking it might be easier to package chromium and then build inox on top of it <efraim>expecially since it seems to be a set of patches on top of chromium <efraim>aqbanking update failed on hydra due to hash mismatch <ng0>dvc ontinued the patch <ng0>if it was dvc.. and it fialed just at some point <ng0>patches applied fine <dvc>I got past the python races <ng0>I moved it to my packages path repository to maybe fix it, but i'm working on so much it's not good <ng0>efraim: it's probably a good idea <dvc>efraim: do you want me to push my patches somewhere if you'd like to continue? It should be easy to uncomment the addition of the inox patches if you want to build chromium first <ng0>and I'm not sure about wificurse.. like, I have no problem with such applications, but I also have the intention to create another live-system after I'm done with the first, focused on pen-testing, so as long as other distros can ship such applications for educational and professional purposes I see no problem <dvc>actually it's only one patch and I noted in the commit message that it builds to 6634/18623 <ng0>I think it's not so bad that I've so much work in progress/stuck, but it keeps me from working on other ideas. After the lispf4 missery I figured it would be great to have an actual interlisp implementation which can also get features from siemens-interlisp like I need it. But to not rewrite too much I wonder if that's something guile can be extended with? I haven't read into the srfi's and all, just <ng0>theoretical thoughts <ng0>I think you need some experience in the system to finish big packages like browsers, and long breaks. I managed to get past a couple of blockers in palemoon and now it's building <ng0>in default layout systems like Gentoo it was easier <ng0>or maybe easier because everything was almost all the time already packaged <dvc>ng0: definitely. you have to be willing to dive in to the code base, have a large pain threshold and patience with extremely long feedback loops... <ng0>but I am really stuck with mozjs38.. it feels like overkill perfectionism to be the only one who unbundles most of the packages out of 0ad, but I want it this way <rekado>I’m trying the prosody service now, but I’m stuck trying to create accounts with prosodyctl <rekado>I created the directory and changed ownership to the prosody user and group, but prosodyctl still gets a “permission denied” error <ng0>maybe we also could package FossaMail, it's what palemoon is to firefox, for thunderbird. I can't use such software these days, but maybe it's possible. If it uses the same friggn build system I'd also consider writing a short build system just to make the package look more clean <ng0>dvc: are there any changes in the inox compared to what you have on your github repository? <snape>rekado: do you use prosodyctl as root? <rekado>installing strace now to see what’s going on <rekado>ng0: what’s the deal with interlisp? Is it used by anyone at all today? Why not Common Lisp? <ng0>because I have this book, and rather than rewriting it completely to be used with Guile or common lisp, I want to use as little modifications as possible to the lecture code base (which is long). and curiosity.. as there's a free software lisp of almost any kind but not interlisp <ng0>the book uses SIEMENS-INTERLISP (all caps yeah), which is a bit different from what lispf4 has too. <ng0>siemens-interlisp also had the one paren to close them all, and things like that <ng0>You could say curiosity in history, and the lack of readable free sourcecode were the trigger. <ng0>so far it looks as if palemoon is really building, but the build took 2 - 23 hours back on gentoo depending on machines. <dvc>ng0: yep, this is the latest. It includes disabling parallel build. <dvc>if you'd like I can push an update to github... <dvc>so updated it incase someone else encounters it <ng0>why do people complain and joke about emacs when there is chrom{e,ium}, firefox, and similar big things.. what goes into a web browser today, incredible <snape>rekado: it's weird you had to create and chown the prosody data dir (/var/lib/prosody). It's something 'prosody-activation' should do <rekado>the problem seems to be this: /var/lib cannot be accessed by anyone but root. <snape>is that the standard GuixSD permission? <ng0>here it is drwxr-xr-x 8 root root 4096 Dec 24 11:51 lib/ <snape>thanks ng0, so it looks rekado's permissions were wrong <ng0>meh palemoon failed, but progress until 35:47.96 is good <rekado>Only used guix system reconfigure <ng0>how far away are we from kde on guixsd? <thomasd>ng0: there's a lot left to be packaged :-) I believe hgoebel is/was preparing a mass import of packages (which would still require some manual adjustments) <ng0>the guixsd defaults are very bash centric, right? I would have to make adjustments to exported variables etc if I made tcsh the login/default shell, or has someone tried this before? <rekado>prosody when started via the service doesn’t provide encryption support. Maybe that’s a problem with finding luasec. <snape>rekado: did you generate the certificates? <rekado>oh, my bad: I provided the wrong path to the key. <adfeno>I'm trying to make recipe for abbaye (from their current repository, not from Google Code)... <adfeno>And, for some reason, the build phase seems to try to call "cc" command, instead of "gcc". <ng0>(setenv "CC=gcc") or you can pass it to the make-flags <adfeno>However, the only thing I do in the Makefile, is: (substitute* "Makefile" (("usr") (assoc-ref outputs "out"))) <jmd>ng0: That should be (setenv "CC" "gcc") <ng0>in the make-flags it should be "CC=gcc" ..mixed up <rekado>excellent work with the prosody service! It’s all working well. <adfeno>Can you all remind me again: (1) What's the difference between "'" (apostrophe) and "`" (accent) in Guile? (2) What's the difference between using (assoc-ref %output ...) and (assoc-ref output ...) ? <rekado>adfeno: the first is “quote”, the second is “quasiquote” <thomasd>% has no special meaning, it's just a convention for certain variables in guix <rekado>adfeno: quote is used to … quote an expression <rekado>adfeno: quasiquote is a toggle switch that is in “data mode” by default <rekado>adfeno: you can switch it to “code mode” by flipping it down (“,”) <thomasd>ha, never though of the switch flipping analogy :) <rekado>adfeno: example: `(foo bar ,(string->symbol "baz")) <snape>ACTION breathes a huge sigh of relief <rekado>adfeno: this gives you a quoted list containing three symbols: foo, bar, and baz. <rekado>adfeno: flipping the switch down before “(string->symbol …)” ensures that the expression is evaluated as code. <adfeno>I find it odd that some recipes in guix use "`" (accent), and others use "'". <adfeno>So, when using ' (apostrophe), I'm unable to use ,(...) inside? <thomasd>what helped me was playing around with some different quoting expressions at the repl <adfeno>Now, let's try building abbaye again : ) <adfeno>It seems to be expecting sdl-config... <adfeno>Does one of our sdl packages provide an "sdl-config" or "sdl2-config" command? <efraim>sdl and sdl-union both provide sdl-config <adfeno>So I guess I forgot that our sdl2 reffers to version 1-and-something <jmd>When will the goddam staging branch get finally merged? <adfeno>How do I tell Guix that I want sdl-union to use version 2 of the default SDL packages provided? <buenouanq>sounds like you're maybe doing something more complex though <ng0>it should be just ("sdl-union-2" ,sdl-union-2) if you want it in inputs and this version exist <jmd>adfeno: ` is the quasi-quote whereas ' is a quote <jmd>` prevents everything in the parens being evaluated (unless unquoted) whereas ' prevents only the first level. <adfeno>Should I call version 2 of sdl-union with "sdl-union@2" or "sdl-union-2"? <adfeno>Sorry, I mean, when inserting the package as an input for another <adfeno>Hm... sdl-union-2 is unbound variable... <davexunit>there is no sdl-union-2 variable in this module <davexunit>there is a procedure named sdl-union, though. <adfeno>davexunit: Indeed... There isn't. <davexunit>which you can use to make the union package that you require <adfeno>But, as you said that procedure exists. <adfeno>Hm... let's, this should be OK according to what the game requires: ("sdl-union" ,(sdl-union (list sdl2 sdl2-image sdl2-mixer))) <davexunit>civodul: so last night, as I was on the brink of giving up, I got a dynamically linked guile booting on ubuntu (sans /gnu/store) and was able to load guile-sdl2 <civodul>davexunit: congrats, quite an achievement! <civodul>so it doesn't have any RUNPATH, right? <davexunit>I had to resort to using patchelf to modify the program interpreter, though. <davexunit>I can see that our gcc package makes some changes to set the right dynamic-linker <davexunit>but to avoid the patchelf hack I guess I need to recompile gcc with some other dynamic linker set <davexunit>I feel a bit lost there. if I remove the code that modifies that stuff then gcc fails to build. <civodul>though one could achieve the same result without the hack, probably <civodul>not necessarily difficult but always takes a lot of time because you never know if you got it right until it's done building for an hour ;-) <davexunit>I'm not modifying any core gcc package, btw. this is a separate gcc used specifically for this crazy project. <davexunit>from what I can see, the program interpreter should be /lib64/ld-linux-x86-64.so.2 on mainstream distros <davexunit>but if I make a gcc that produces such binaries, I'll need to disable test suites and stuff because there's no way the resulting binary is going to run on a guixsd system <davexunit>unless it's possible for me to create /lib64 in the build container <ng0>oh, palemoon almost builds <ng0>need to leave now :) <nliadm>this is why I dislike dynamic linking :/ A bunch of assumptions that are completely inscrutable when they don't hold <davexunit>please let's bring this flamewar topic to this channel <davexunit>it's only when you're doing weird shit like I'm doing that dynamic linking can seem odd. <davexunit>trying to make a dynamically linked binary that runs on many distros is admittedly a stupid thing to do. <civodul>davexunit: doesn't Debian use /lib/x86_64-linux-gnu or similar? <nliadm>I've been there, try to make new binaries to run on old systems I couldn't install compilers on, etc etc <thomasd>I have found the "LSB Application Checker" very useful for that: it contains a database of which distributions ship which version of standard libraries, and will check your binaries for undefined symbols against that database <nliadm>The LSB Download page is temporarily down, and will be made available shortly. <davexunit>civodul: I could just include ld-whatever.so in my bundle and run the binary that way, no? <thomasd>unfortunately it seems like it hasn't been udated for a long time (maybe they gave up on it now that we've entered the container era?), but if your app satisfies the constraints of those distributions (~3-4 years ago) it will run anywhere :-) <davexunit>if I could assume unprivileged user namespaces everywhere this would be way easier <davexunit>I'd just pack up the closure of the store item and mount it in a new mount namespace <nliadm>I want unprivledged mount namespaces so badly. <davexunit>but as it stands I'm about halfway to making game bundles without any container stuff <civodul>but you need to know its absolute file name <davexunit>I thought the man page said I can start a program with it directly <djLeGit>It seems our `mupdf' does not contain nor generate a pkg-config file <djLeGit>and I'd very much like to package some software that assumes this to be the case. <djLeGit>would it be worthwhile to add a post-install phase, like the Nix project does? <civodul>if upstream does not provide a .pc file, i'd rather not add one <civodul>otherwise Guix users would start assuming that mupdf.pc is something they can rely on <djLeGit>hi civodul; that seems to be a reasonable stance, but I thought I'd ask nevertheless :) <djLeGit>I guess I should patch the source of the package I'm trying to build directly then <civodul>so that package assumes mupdf.pc exists? <civodul>now if every distro provides mupdf.pc, we could follow suite <civodul>it just seems to be more of an upstream decision to me <djLeGit>it's build system is a bit of a mess, but it works wonders for those horrible 2-column papers <djLeGit>this is the monstrosity that the Nix folks came up with <djLeGit>there is a cmake-based thing, but I don't think it works at all <civodul>and it bundles a couple of OCR packages, woow <djLeGit>yes, it includes patches that have to be applied to the /sources/ of its dependencies <djLeGit>I could make some custom versions of its dependencies via (inherit <pkg>), and put the patches for these in the gnu/packages/patches folder. <civodul>yeah, but that's the best you can get, i guess <efraim>Qtscxml passes its tests, but the tests are bundled 3rd party code :/ <lfam>There is a hilarious TIFF file in the windowmaker package that I used to test libtiff: 'WindowMaker-0.95.7/WindowMaker/Icons/BitchX.tiff' <ng0>but bitchx is an irc client.. <ng0>s/is/was/gras/noidea <ng0>libpng with apng support is an maintained patch (or patchset), I will create a new package for this, inheriting from libpng <paroneayea>yeah bitchx is an irc client. had some verrrrry problematic automatic quit messages at least back in the day. <paroneayea>let's see if I can patch up my old gnutls-with-guile-next patch and make it not gnarly enough to warrant a submission to guix proper <lfam>paroneayea: Any progress with python-xdo? <lfam>I'm sorry I broke your password manager. Are you at least able to use the older Guix revision of it? <paroneayea>lfam: yes, the nice thing about guix is that I could leave the old one around :) <lfam>Right, that's a huge benefit. Although it's challenging for new users to install it now <paroneayea>you even gave me a chance to look into it, *I'm* the one who misunderstood and thought things were working and gave you a thumbs up <paroneayea>lfam: hopefully john darrington's graphical installer will help there! <adfeno>"bitchx" really need that bugfix. <janneke>why is it again that guix environment does not update environment variables? <lfam>I'm really impressed with all the work happening in Guix right now. The graphical installer, Btrfs support, tons and tons of new packages <lfam>janneke: What do you mean? In my experience, it works by changing environment variables <janneke>lfam: i do: guix environment -s i686-linux gcc-toolchain <janneke>and then i still have my old C_INCLUDE_PATH, LIBRARY_PATH etc <janneke>so i need a script to find the new profile from PATH, which does get updated <lfam>And you're not setting those variables in your non-login shell initializiation (e.g. bashrc)? <lfam>bashrc will be sourced by `guix environment` so it can clobber things <lfam>Otherwise... I think that davexunit is the expert <janneke>big chance i'm doing something stupid over here... <ng0>is there alread a package which downloads one patch, extracts it, applies it.. that I could look at for inspiration? Otherwise I will construct that myself <ng0>I'm doing libpng-apng, where APNG support is an distributed patch in .gz compression <ng0>and I wonder if there's a short way to add the patch on top of the other libpng patches <ng0>I know the long way, like done in bash, inox, icecat <nliadm>I thoguht the build system cut you off from the network? <ng0>it will still allow fetching sources <ng0>and the patch is a source <lfam>nliadm: The build environment has no network, but Guix can download sources and make them available in the build environment <ng0>I'll just come up with something, the state of palemoon is still so messy that I need to fix much before upstreaming it <nliadm>is there an example of grabbing sources with http basic auth? <nliadm>I know there's support threaded through, I just haven't figured out how to do it yet <lfam>nliadm: You suppose? :) That's how source code is provided to the build... <nliadm>I mean, that you can specify patches as additional sources <lfam>nliadm: Right, it's safe because we record and verify the hashes of the additional sources <nliadm>I remember seeing another answer on the list <nliadm>there's also a specific mention of basic auth in the guix source <lfam>Yeah, ISTR that it's possible, but I can't remember enough to find the discussion now <lfam>nliadm: guix/import/cran.scm line 126 <ng0>should I inherit lipng or libpng/fixed ? libpng/fixed is the replacement for libpng <ng0>well I think I need to libpng/fixed <ng0>when are the patches applied? what's the name of the phase? I imagine I could just extract the gziped patch before the patch gets applied <lfam>ng0: Patches in (source) are applied when building the source derivation, before the package build even starts. <lfam>But, I'm sure you can apply patches in build phases, too <ng0>okay so no way around creating an extra thing I can then move into the patches <lfam>ng0: See the net-tools package definition <ng0>also libpng-apng tells me that we should update our libpng, if .28 is the latest <lfam>ng0: There are grafted security fixes for libpng. We can't update it often because almost the entire distribution depends on it <ng0>I'll just package a fitting version for apng-png now <ng0>if they provide it this way <ng0>I hope I don't have to patch the patch because of our patches <ng0>why is this in inputs wrong: <ng0>+ (uri "mirror://sourceforge/libpng-apng/libpng16/" <ng0>+ version "/libpng-" version "-apng.patch.gz") <ng0>can't I use version and mirror? <ng0>I use the version from libpng <ZombieChicken>Anyone here? Need a little help getting grub to install to a drive via UUID using the system config during an initial install <ZombieChicken>(bootloader (grub-configuration (device (uuid "$UUID")))) doesn't seem to work <ZombieChicken>(please note $UUID is an actual UUID for the drive, and not a variable) <ZombieChicken>apparently I'm looking at the filesytem UUID and not the devices... <jmi2k_>ZombieChicken: determining the block device based on the root filesystem is feasible for you? That's what I do in my portable installation <ZombieChicken>jmi2k_: I'm trying to install GRUB to a seperate drive from the rest of the system (a VM atm). It doesn't seem to like it <jmi2k_>ZombieChicken: root in /dev/sda and grub in /dev/sdb for example? <roptat>hi, I'm trying to run josm (a java application), but I get the error "java.security.InvalidAlgorithmParameterException: the trustAnchors parameter must be non-empty." when it tries to download map data <roptat>I think this is because java doesn't find the certificates, how do I tell it where to find them? <ng0>was there a thread on java development and packaging in the past? I might need this at some point in the future.. anything special on GuixSD to consider when you write java? <civodul>yeah there was a discussion of Java & certs on help-guix <ng0>sorry, I did not mean to derail the question <nliadm>iirc that needs to be set in a jvm properites file <ZombieChicken>I'll probably just send something to help-guix and see what I can get <dvc>civodul: I managed to cross compile all of guix's dependencies, any idea how the guix build system would need to be extended to cross compile guix? <dvc>well I'll look in to it - enough for today :) <jmi2k_>ZombieChicken: I think you could use my solution if you have any filesystem in the destination drive. But why you can't pass the block device path instead? ***jonsger1 is now known as jonsger
<civodul>sneek: later tell dvc normally Guix can already be cross-compiled the normal way <ZombieChicken>jmi2k_: Sorry, I'm only kind of paying attention to this. I'm unaware of your implementation, and right now I'm just installing with --no-grub, then I'll see about hand-installing GRUB and see if that works <jmi2k_>ZombieChicken: np, same here ;) Good luck! <ZombieChicken>jmi2k_: Do you know anything about blocklist errors with grub-install? <jmi2k_>ZombieChicken: no, sorry. I'm not very experienced with grub <rekado>this error bit me again: when doing “guix system reconfigure” after changing the nginx configuration, restarting the nginx service is not sufficient <rekado>it will just reuse the old nginx configuration <rekado>how do I reload the services so that nginx picks up the new configuration generated with “guix system reconfigure”? <roptat>currently you need to stop the service first (herd stop nginx) then reconfigure your system <rekado>roptat: that really did help. Thank you! <ng0>43:26.23 /gnu/store/2qxzpfc8hzd1n91xjwyyrq70wl5f5f0i-bash-4.4.0/bin/bash: hostname: command not found do we have something providing hostname? <ng0>i mean we have, but what? <ng0>I wonder why palemoon build needs it <civodul>rekado, roptat: yeah i often forget to stop the relevant services before reconfigure <davexunit>nginx in particular is a service that needs to be able to hook into something when it's service implementation changes <davexunit>for zero downtime updates we need to send nginx the USR1 signal (I think) so that it reloads config without shutting donw <rekado>I always expect “herd reload nginx” to do the right thing, but it doesn’t do anything :) <davexunit>and then for upgrades of the nginx binary we need to send a different signal <davexunit>and nginx will replace itself without dropping any traffic <civodul>yeah nginx is very fancy, but that relies on being able to find the config file and binary at a fixed location <davexunit>then there's other services, like mysql, which you want to be able to say "don't you dare ever restart this automatically" <civodul>right, it can't guess where the new file is :-) <davexunit>I would really like to have those features available <davexunit>we can symlink the config and binary to somewhere in /run or /etc or whatever <civodul>yes, we could do something like that <civodul>in fact, that's quite easy to do and we don't need shepherd's help for that <civodul>killall -USR1 nginx or whichever signal that is <sturm>I'm looking to modifying the config of `mysql-service` to change `max_allowed_packet`. What's the Guix way of making this change? From "info:guix#Database Services" it looks like I can't do this on the `mysql-configuration` object.