IRC channel logs
2023-06-03.log
back to list of logs
<Fare>Hi. I'm trying to build cryptsetup-static-2.3.7 on arm64, and its ./configure fails with: configure: error: You need the gcrypt library. <Fare>how do I even start to fix that? <Fare>ACTION tries to update his channels-lock.scm and pray <oriansj>You know something is missing if praying is required when using a declarative package manager. <oriansj>with perhaps a way to collect and report on which versions work and which don't (possibly on which architectures) <Fare>Often, upgrading the channels solve an issue, and introduces another :-/ <Fare>I love Guix so much in theory, but in practice, it's so much harder to use than NixOS, despite the better language. <Fare>Updating the channels didn't help :-( <|F9|>Fare: I loved guix, but it's too slow <|F9|>I am not sure, maybe it's my side problem. <|F9|>But 5-6 minutes to install little package?! <|F9|>Like (random system just to compare) on FreeBSD it would take 7 seconds <Fare>|F9|, you need to add substituters <|F9|>Anyway, you all can dislike me now <Fare>Nix runs on OS X... though as of late, the installation isn't seamless anymore. <|F9|>Thanks for link, I may try <|F9|>A bit offtopic, but overall, how bad is OS X? <Fare>I think the one that was missing from my config might have been bordeaux.guix.gnu.org <Fare>|F9|, evil enough. But they have the fastest / lowest power hardware. <|F9|>Yeah, just OS X is so easy and simple to use, it's just working. <|F9|>You don't have to spend 10 hours to learn it <Fare>You just have to work 20 extra hours to afford it. <Fare>and then become Apple's bitch. <littleme[m]>π»oh yeah, and you became a small part of dear osx, rm -rf money_pride_control <littleme[m]>and ifconfig lots of unknow interface π hope nix at osx save your soul. <littleme[m]>Maybe you will feel pride about yourself that you can afford double even tripple price to use a machine that just work like others. π» dear, and the software actually suck if there is no electron-tech <littleme[m]>unless you choice M1, it's not cheap, at least it show the design of good power manage to CPU. <|F9|>Fare yeah apple stuff is very expensive <|F9|>littleme[m] even considering I hate apple pretty much... <|F9|>Nobody in the world makes so great quality hardware like they. <|F9|>like, software, it's great, but non-free, so it makes it bad. <|F9|>But hardware, it's awesome. <|F9|>I always can install some Trisquel or FreeBSD or even 9front to macbook <littleme[m]>nope, π»it's not such pretty without m1, it stuck so long. there is Thinkpad X1,dell xps, so much choice now. it did guide the design before, but not today <|F9|>littleme[m] do you know what design is? <|F9|>I know what design is, that's why I choose apple's hardware. <|F9|>Also you can libreboot macbook <|F9|>And (old ones) use open firmware <|F9|>I was talking about case, screens, etc. <|F9|>Yes, old hardware is slow, but it's great. <|F9|>And you can run free software on it. <|F9|>Saying that even though I dislike GNU/Linux a lot and slowly purging it everywhere. <littleme[m]>screens π it's 2023 today, you can't find a worse screen with a mac-price-level <|F9|>powerbook g4 vs any of my laptops and any of my monitors <|F9|>I will able to test ThinkPad X1 Gen 5 today <littleme[m]>π»it's such cheap π»π»π» i got X1 Gen 3 at 300$, come on π». <|F9|>Don't use so much emojies, it looks silly. <|F9|>Congrats if you are rich, but not everyone are. <Fare>|F9| if you can't get the top of this year... you can still get the top of a few years back. <|F9|>Don't compare what can't be compared. <|F9|>All I know about X1 is that it have shitty design, and people say it's great overall (apart from design and chinese keyboard) <littleme[m]>but i failed to configure the audio driver at Guix in X1, but i hardly play sound <Fare>$1000 is so low compared to what machines used to cost back in the day... and that's not even accounting for all the inflation! <|F9|>I'm audiophile, sound is very important to me. <|F9|>Unless you live in Ukraine like I do <|F9|>And you have 57 usd per month free to use <|F9|>Also, about software, it mostly shit everywhere. <|F9|>Windows is unusable garbage, GNU/Linux is usable but still overbloated garbage <Fare>The ruscists destroyed the museum of computing in Mariupol. I can't forgive that. <|F9|>I got explosion 500 meters away from me <|F9|>They reached kids clinic <VesselWave>Is there a way to install mu4e as a guix package? Or how get it the guix way? <davidl_>what is required for a package's info file to be installed correctly when you make a package using the gnu-build-system? <janneke>jpoiret, civodul: i managed to build "hello" natively! <futurile>davidl_: are you sure it's not being 'installed' (copied to the right location), versus your profile not finding it? AFAIK if the underlying 'make install' command works then the docs/info should be installed correctly <janneke>civodul: you asked about all the new failing coreutils tests on the hurd and debian <janneke>i didn't find patches or anything before, but i just found how they do it <janneke>in coreutils's debian/rules there is: <janneke>it seems debian just skips all coreutils tests on all platforms, iirtc <janneke>ACTION built coreutils on debian-hurd using dpkg tools and found it went from 'build' straight into 'install' <civodul>these are failures of the Gnulib tests i guess? <civodul>to some extent it's good because it avoids discovering problems down the road <civodul>OTOH, they're sometimes too picky, like testing things that aren't mandated by the interfaces <janneke>civodul: there are still 6 gnulib tests that fail (one less), but no, all new failures are coreutils' own tests <janneke>civodul: i haven't looked into the test suite, i would hope that they have basic tests, and then picky ones <janneke>ACTION wonders if such feedback ever reaches upstream: your tests don't work for us, we turn them off <davidl_>futurile: basically, Im writing a Makefile right now for one of my own projects, trying to make things work as smoothly as possible with Guix' gnu-build-system. I am new to using make though, so Im not sure how .info files are normally handled. <davidl_>I have a build directory where I put all files, like build/bin, build/share/project, build/share/info, I just wonder how things are handled with info files so I can avoid trial and error with it. <lilyp>unless you have a good reason not to, I'd suggest using autotools <janneke>ACTION pushes a (their) wip-hurd branch with the two pending patch sets (that need some cosmetic changes) and the to-be-sent commencement one, and some goodies <jpoiret>i've been derailed by misc stuff this morning (wanted to upgrade b4, but it has a new dep that needs to be upgraded, and it doesn't generate its own doc) <janneke>i'll be afk prolly till tomorrow in a few <jpoiret>what do I do with patches that overall solve something, but have a bunch of mistakes that I fixed locally? should I just push them with all my changes? <jpoiret>janneke: are you using the new tagged mig btw? <jpoiret>I haven't followed too closely, my own local branch has it <jpoiret>if gnulib's tests fail, won't we have problems down the line with other packages using gnulib? <janneke>i'll prepare a 'commencement update / native build hello' patch set from wip-hurd once both our pending patch sets are pushed <janneke>that could be, but goes on a package-by-package basis, i guess <janneke>"<jpoiret> what do I do with patches" -- i don't get what you're saying, are those patches with mistakes already pushed? <jpoiret>i was wondering if it was fine to push heavily modified patches under the original author's name, since you've changed their patch quite a lot <janneke>if the mistakes are obvious/trivial, no need to send another review round, i would guess <janneke>if i change anything more than cosmetics (it's not black and white), i usually add a co-authored-by <janneke>jpoiret: feel free to revert + push-changed any commits to wip-hurd <janneke>(or reset the branch entirely, preferrably with a heads-up notice) <jpoiret>yeah, I'm wary of adding new branches because I like rewriting the history, don't know how to handle that very well when other people are also using it <janneke>yeah, it can get tricky if you don't align <janneke>i usually keep adding squash! commits until we agree to squash them all, and reset <davidl_>lilyp: ok thanks for the suggestion. My projects are not C. They are mainly bash and python projects, but still for at least Bash it might be useful. I just want to make it into an easy and standard build process. <ChocolettePalett>You could look how they do it in other projects, but I can't remember any project that uses info files <jpoiret>ChocolettePalett: do you mean info manuals? <jpoiret>lots of projects use info manuals, I suggest checking out your `info` do have a look at some examples <jpoiret>they're less popular then man pages for sure though <lilyp>davidl_ for python, prefer the pep-whatever build systems, there are some useful ones in guix <lilyp>(assuming python is the only language in the repo) <dthompson>jpoiret: I totally missed your question yesterday. I was having issues where a guix pack binary wasn't running correctly because it couldn't find glibc locales. I was able to solve this problem by using a wrapper script that set GUIX_LOCPATH appropriately. <jpoiret>petition to add forceinbodyfrom=true to our git config <spacecadet[m]>Hello guix, I'm in the unfortunate situation of needing to run guix shell inside a guix system vm. Even if I wanted to give the vm r/w access to the store (I wouldn't) I don't think that's possible to do safely. Does anyone know a workaround here? <civodul>spacecadet[m]: hi! you could use a full-blown VM image ("guix system image -t qcow2"), and there you could run guix-daemon and all <davidl_>ChocolettePalett: me neither, other than that I know that with guix I can use info guix :-) I checked it out but didn't find it informative. I eventually just looked at the gnu-build-system in the guix source, and I think it'll work out fine if I just have a make install phase that installs some info files to /usr/local/share/info, however, it remains to be seen! I will just have to do trial and error soon. <lilyp>whatever you do, just make sure to honour $prefix and you ought to be fine <davidl_>yeah seems like it. Ill probably do a preliminary regular configure and Makefile scripts before sswitching to configure.ac and Makefile.am <VesselWave>Hi! How to send a guix patch with mu4e the most efficient way? <VesselWave>TheSkylarverncc[: What file do you want to modify? <jpoiret>guix records are a ticking time-bomb <patrickt>Hello, sorry if this was already asked and addressed (that last few days have been pretty draining). Trying to set up Guix on a new system but the 1.4.0 installer keeps failing with 3 different internet connections. I ran guix pull in command line, which reported back guix pull: error: Git error: the SSL cetificate is invalid. Help is immensely appreciated. I am also eager to help make the installer more streamlined (would be nice if it could <patrickt>automate more of the disk partitioning stuff) <patrickt>By failing, I mean, it says the substitutes servers are inaccessible <VesselWave>TheSkylarverncc[: Sorry I haven't really used guix home <VesselWave>TheSkylarverncc[: plain-file reqiures string, computed-file requires gexp. I don't know how to make gexp that outputs string <patrickt>evilsetg: Ah snap, good catch, thank you so much for your help! <mekeor[m]>note that you need to have required the -mu4e feature <mekeor[m]>btw, when you want to apply patches from an email with mu4e, i recommend: (add-to-list 'mu4e-view-actions '("apply git patch" . mu4e-action-git-apply-mbox)) <civodul>jpoiret: "guix records are a ticking time-bomb", what makes you say so? <civodul>(i can see ticking time bombs in Guix, but not this one :-)) <jpoiret>civodul: the macro foo is bound to blow up in our faces (the match-record patchset is one example) <jpoiret>see #guile for me desperately trying to debug a related problem <jpoiret>i don't know how many people understand how guix records work, but i don't expect it to be more than a handful <patched[m]>Hey, irc logs since january don't seem indexed, so I'm asking again... <patched[m]>How can I make libstdc++ available in guix shell? <civodul>jpoiret: aaah, i see what you mean :-) <jpoiret>patched[m]: add `-e '(list (@ (gnu packages commencement) gcc-final) "lib")'` <civodul>it's heavy-handed macrology, and it's become hairy <jpoiret>yes, especially if we want better error-reporting <civodul>for that we should look at the other syntax-* that Racket has, perhaps <jpoiret>i'm thinking that we should drop "records: match-record: Display more helpful field-not-found error." <jpoiret>that's the source of the (current) problems of the patchset, the rest LGTM <mekeor[m]>patched: wildly guessing, try "gcc-toolchain:static" xD <civodul>i thought the patch set was mostly ready? <jpoiret>no, that code causes `gnu/services/networking.scm` to not compile anymore <jpoiret>because match-record is used with a body with an ellipsis in it <civodul>so we need to investigate what you posted in #guile <jpoiret>even from a very close inspection no one could've guessed that this would bite. and i was about to push it! <patched[m]>It's not in gcc-toolchain :^} I used gcc before the last core update, when it was removed <jpoiret>i can push the patchset without that offending one <mekeor[m]>ACTION wonders if the libera-matrix bridges works again <jpoiret>mekeor[m] patched[m]: can you read our messages? <civodul>in "records: match-record: Display more helpful field-not-found error.", the #,s bit at the bottom looks suspicious to me <jpoiret>no, here you want to insert the whole expression, and it's not bound by syntax-case/rules <VesselWave>mekeor[m]: Thank you so much! What patches are you appllied to your git-email fork, why I should use it? <jpoiret>i reduced the bug to the MWE i sent in #guile, which i think is a guile bug i think <jpoiret>(here's me repeating things in the same message, not looking good) <jpoiret>civodul: I just thought I'd have a go at that patchset today while pushing some other stuff because it's been hanging around for a while. <jpoiret>but yeah, I don't think there's any rush to get that working <civodul>if dropping that one patch works, you can do that <patched[m]>I have tried all the outputs. That's why I went to the chat last time I asked here <jpoiret>i'll need to modify the next patches but yeah, it'll work <jpoiret>VesselWave: the matrix-irc bridge is down <VesselWave>I've never heard that guix irc is bridged with matrix, lol <jpoiret>VesselWave: people with the [m] are on matrix <jpoiret>the bridge has been having issues for the past few weeks <jpoiret>civodul: i think the workaround of making it syntax->datum then back when raising the error will work, but it's quite ugly <jpoiret>okay, I have it working, i'll try it out and push that <mekeor[m]>patched: alright. maybe asking on help-guix mailing-list might be worth it <patched[m]>mekeor @mekeor:matrix.org: I recall somebody sharing a hack to make gcc available in a shell, by passing an expression to guix shell -e <mekeor[m]>oh, it's even more confusing because *some* messages do cross the bridge from libera to matrix. thanks, val! <jpoiret>civodul: so should we simply remove the syntax-case for match-record? I don't think that's the one thing causing problems here. Would you want to replace it with syntax-rules?= <mekeor>emacs-ox-pandoc has (mistakenly?) been packaged as version 20180510. the latest version is "2.0". how to treat version decrease? <civodul>jpoiret: it is syntax-rules currently on master, before this patch set <civodul>and then, without the #,s bit, there's no risk of introducing a user-supplied ellipsis down the road <jpoiret>oh, you're right. Here i guess the objective *was* to get the whole form in the error message <jpoiret>but yeah, might be too big, and we only want the location <jpoiret>so let's just not apply that one patch <jpoiret>i wonder why the error message doesn't have the proper location though <ngz>sneek later tell rekado I kinda start to understand the hyphenation issue we're facing in modular TeX Live. It seems that texlive-babel must be among the TEXINPUTS when latex format is generated. IOW texlive-babel must an input for texlive-latex-base, which creates a circular dependency. <civodul>jpoiret: for me it gets the right location, but that's the location of the whole match-record form; we'd prefer that of (fields ...) or similar <jpoiret>i don't think giule can give us the location of the subforms unfortunately <jpoiret>i'm very surprised by this but even the simplest syntax-rules macro with a syntax-violation on a bound piece of syntax can't report the location of it <ngz>sneek later tell rekado circular dependency that is solved by building texlive-babel with pdftex engine. This does bode wellβ¦ <civodul>jpoiret: without the patch is fine; i do get the right location with the syntax-rules version i posted though <civodul>we have tests for instance in tests/guix-system.sh that check for location in error reports <janneke>jpoiret: shouldn't these be (libc-for-target)? <jpoiret>civodul: ???? now that abracadabra trick i really don't understand <mirai>Any gexperts know how to read the built (mixed-text-file ...) from within the same source? I'd like to write a guix test that checks if the serializing procedure really writes the correct output <https://paste.centos.org/view/6093867e> <jpoiret>how does adding some random piece of syntax that you don't use in the end add the location? <mirai>I'd like to essentially automate ",use (guix) ,build (mixed-text-file ...)" ; "cat <the resulting path>" <civodul>jpoiret: the piece of syntax we introduce inherits source location from the original form <civodul>we don't actually care about what the content of that syntax object, we only care about its location <jpoiret>civodul: yes, but even if we don't pass it to syntax-violation in the end it still gives the right location, which is what I don't understand <jpoiret>i don't think we can get more precise than the match-record form location, but I guess that's fine, it's still logically close enough <civodul>we pass it to syntax-violation: it's the 's*' there <jpoiret>heh, no need to pass all of that, we can just re-use the s from the lambda of lookup-field <jpoiret>we just need to rewrite it to not show the fom <mirai>jpoiret: what's the "format helloes" for? <jpoiret>mirai: ah, you can replace test-gexp by any file-like <mirai>related, is it possible to sidestep "writing to a file"? That is, to get as a string what #~(....) will be "rendered" as <jpoiret>or you can also have a look at (guix monad-repl) <mirai>gexp->sexp isn't recursive iirc <mirai>(eval (gexp->approximate-sexp ...) (current-module)) is out of question when it contains nested gexps <lilyp>pretty sure you did something wrong when you have a gexp in (string-append ...) <mirai>jpoiret: nope, Z is #~(string-append #~(format #f ....) #~(format ....)) <lilyp>yeah, those #~(format ...)s are superfluous <jpoiret>mirai: i would expect gexp->sexp to work <mirai>lilyp: nested gexps naturally occur in define-configuration <lilyp>don't they get intermixed with #$ tho? <mirai>(gexp->approximate-sexp ...) removes those <mirai>otherwise it wouldn't even pass the ,build in the "original" paste <mirai>(if the gexp wasn't intermixed with #$) <lilyp>Like, a define-configuration with one boring type (e.g. an integer) field and a serializer <lilyp>we have different opinions on what's minimal then <mirai>truly minimal would be beginning at 163, what would be the proper way to write a guix test that ensures that the serialization gives the expected output <lilyp>that's not minimal, it depends on all the fluff that was before