IRC channel logs
2023-01-03.log
back to list of logs
<nckx>You used to be able to scrape by without DKIM if you made sure not to send (to Gmail) over IPv6. <nckx>And here I was being cautious in my wording ☺ <nckx>I still strongly recommend it. <nckx>But it's not an instaspam signal, nor will setting up DKIM magically deliver your mails to everyone's inbox. <nckx>ACTION assumes you have SPF/DMARC, we're not animals. <KarlJoad>Right now, I have nothing. Just trying to figure out how to best keep my secrets secret without needing to jump through a ton of work for GPG encrypted files to read my password to send emails. <KarlJoad>Automated emails from my Cuirass instance. <nckx>I don't know what kind of server this is, but do you need secrets? <nckx>Like if you're just sending mail from Cuirass to your own Gmail account, I'm not sure you do. <KarlJoad>It's a personal server sitting next to my desktop. I am using this as an experiment really. I can put the password in plain-text if I need to though. <nckx>Forgive my ignorance, but which password? <KarlJoad>I already have it working for a plain-text password and a generated send-mail config file. But I am now looking at ways to not have that in plain-text. <bjc>if you want it automatic with no user intervention, you'll need the key somewhere <bjc>one option would be to read the key to decrypt the password as root, then drop privs so the keyfile is no longer readable <nckx>I thought you had or were building an opensmtpd server. Now you're (also?) submitting mail to Gmail. I'm confused. <bjc>i think you can have opensmtpd act as a relay so it'll take mail and forward it on to an smtp submit port, which will require auth <nckx>Yes, but there is no way to bring DKIM into that story. <nckx>Submitting (as in port 587) to Gmail is totally fine; it just isn't setting up a mail server. <KarlJoad>This is a Cuirass server. I am trying to get emails to send on failures. I was recommended to use opensmtpd as a relay, that's about it. I already have it working though. <KarlJoad>Just as an experiment, I want to have the password garbled/encrypted somehow. <nckx>Yeah, I think you were misled, DKIM is completely irrelevant here. <nckx>As are SPF/DMARC/rDNS any other scary stuff you might have heard of, by the way. Even opensmtpd is technically superfluous, but using it as a relay does have advantages. <nckx>I'm sorry, I got all exited that you were spinning up a mail server before bed time. I do that. <KarlJoad>I mean... I may eventually spin up my own mail server, but not right now. First up is to get emails flowing from my Cuirass server to my address. I can do it with a normal send-mail config, but I have been recommended opensmtpd instead. <nckx>Just put your password in a file that's only readable by smtpd (and root). <nckx>It's not your main password. <Kolev>I'd like a volume that keeps all my keys and acts as a physical token. <nckx>ACTION notices ‘App Passwords can only be used with accounts that have 2-Step Verification turned on.’ Dunno if that's you. <nckx>I'm not a Gmail user myself; supported it for a while (under duress, I swear) but that was all business stuff with 2FA. <nckx>What a silly restriction. <bjc>it's relatively new. i think they're just trying to push 2fa, and this is a stick they can use <KarlJoad>Yeah. That is something I should look into. But for now it only complicates things. I guess I will set up opensmtpd with the plain-text password and re-write the mailer script. <KarlJoad>nckx: How do you get those "status messages" to happen? <Kolev>I don't like how 2FA is phone-based. <nckx>bjc: Literal lol; I was thinking just that but didn't want to say it. <Kolev>nckx: At work, I have a smartcard, but with Google, they use SMS and phones. <bjc>pretty much all my 2fa is whatever that otp spec is, and handled by keepass, including google's <nckx>KarlJoad: Not sure what you mean by status messages. <KarlJoad>The "nckx notices 'App Passwords...'" was what I meant. <bjc>sms 2fa is just a fallback at most places these days, ime <nckx>I use my laptop as second factor, works fine. Is it strictly less secure than a second piece of hardware? Yes, but that is completely orthogonal to ‘2FA’. <nckx>/me illustrates the trick. <Kolev>I might be migrating from KeePass to Password Store. <KarlJoad>ACTION broke their currently booted system, again... <bjc>i don't think i'll ever migrate from keepass. it works on everything and is easy to sync with dropbox/syncthing/whatever <nckx>bjc: I use oath-toolkit, and I haven't encountered a site that didn't work with it, so I guess that's the same standard. <nckx>ACTION knows too many RFCs by name. <bjc>i have forgotten most of them, thankfully <bjc>correllation, etc, but the fewer rfcs i remember, the better my life has been <nckx>Kolev: Unfortunately, there are some stupid implementations of 2FA (like implementing it as SMS), but there's nothing phone-centric about 2FA itself. <Kolev>nckx: Sure. I'm just saying the average person uses SMS for it. Nobody has a smartcard for their personal laptop. 😂 <nckx>I think that might be cultural/geographical. <Kolev>Really? Every web service we use texts you a number that you have to enter after you log in. <nckx>The only area where that's remotely true here is banks, because they can afford to live in 2008. <Kolev>nckx: So Europeans have smartcards for their email? <nckx>Kolev: For each log-in? That's nuts. <KarlJoad>That was strange. When I switched generation, all my SUID programs did not get repopulated in /run/setuid-programs, so I had to reboot. <bjc>Kolev: you don't need a smartcard. there are plenty of 2fa applications out there. google even makes their own (which will work with most services, since it uses the above rfc) <paul_>Kolev: yeah, TOTP support seems to be pretty common (for sites that offer 2FA) <paul_>(unless its a bank or medical provider, which you would hope would be the most secure of web platforms, but regulatory beaurocracy does what it will) <nckx>Kolev: Maybe ‘2FA’ has a different meaning where you live/work, but yeah, same as bjc. I don't claim to speak for all of Europe; but in my circles, anyone technical will react to mandatory SMS reception as quaintly antiquated, high-maintenance, and insecure, and anyone nontechnical merely as antiquated. <bjc>if a website asks me for a phone number, i don't use it <nckx>ACTION doesn't have a phone number, so it's kind of a deal breaker. <Kolev>nckx: 2FA in the U.S. is either smartcard or SMS in most cases. <bjc>i'm dead positive it's just a data collection device <nckx>Kolev: Interesting. Here, our national ID card is a smartcard and is used for Web auth, but that's only for government services. It is for *all* government services, though, which is nice. <nckx>Kolev: Interesting II. Here, while few will know the term ‘TOTP’, the workflow is standard. <nckx>‘I need to open the log-in app.’ <Kolev>nckx: Smartcards are used at workplaces. SMS 2FA is used for all personal accounts. <paul_>'provide me the magic 6 digits oh great time device' <Kolev>Maybe I should start using TOTP. <nckx>Kolev: I love how the US does things that are considered a decade out of date here, and vice versa. <Kolev>nckx: But at least Americans can still use the anonymous payment system called cash. <Kolev>I hear cash cannot be used in Europe. <nckx>Yep, that's a good counterexample. <nckx>Uh, no, that's not true. <KarlJoad>What is the right way to specify remote file systems? I did not find too much useful info in the manual. <Kolev>nckx: I'm not saying cash is illegal in Europe, but too many places will demand card payment only. <nckx>Here, while I can tell you the name of a coffee place that does so, I can do so because it's exceptional. <nckx>And I'm sure some place in SF does the same. (OK, maybe a stretch.) <nckx>rekado uses NFS for $stuff but I don't know if it's in that way. <nckx>ACTION → 😴💤, nice chatting to y'all. <KarlJoad>nckx: I'm using CIFS for now, but may move to NFS if that is simpler. <KarlJoad>nckx: Thanks! rekado: Could I see your configs for remote NFS mounting? I want to mount my NAS's exported directories on my desktop. <jts>is there a mechanism for inheriting an operating-system reference and modifying only certain fields? I have two machines I'd like to have the same configuration with different hostnames and, of course, file-system references <iyzsong>jts: yes, see 'os-with-uboot' in gnu/system/install.scm, one just need to use (inherit ...), like other records. <johnabs[m]>So yesterday I found a weird bug with inkscape where I couldn't open png files from the Open dialogue box...Nothing worked to fix it, until I opened a pdf? Then I opened a JPEG, and now I can open everything...IDK why, but it works now, lol. <johnabs[m]>And weirdly, this persists after I close the program <ConvolutedSquare>Just in case anyone is booting off a USB and mounting over /gnu/store, here's some magic to copy all kernel stuff used by GRUB: sudo cp -a (grep /boot/grub/grub.cfg -Poe '\\/gnu\\/store\\/[^;\\s\\/]*') /.gnu/store/ <FriendFX>Hi Guix, I have a question about Intel microcode, I am guessing this is off-topic here? <sner>FriendFX: if you're looking for high-level info i'm happy to help, but nonguix would probably be the best place if your aim is more pragmatic <rekado>efraim: I wonder if maybe icedtea itself is subtly broken on aarch64, given the fact that java-testng fails horribly. <civodul>rekado: howdy! i'm looking at the 'guix refresh' breakage <efraim>rekado: Just made it to my machine again. gcc-7 worked for openjdk 9 and 10, gcc-10 is fine for openjdk 11, openjdk12 failed to build <rekado>civodul: I just sent a new, simpler patch to work around this. <rekado>efraim: haven’t tried anything beyond 9, but it did work all right with just ("gcc" ,gcc-7) added to the native inputs. <rekado>but I wonder if the JDK might not be miscompiling things on aarch64 <rekado>I’m scared by the many weird test failures in java-testng <efraim>the patchset said 8.3+ for gcc. I've been adding gcc-toolchain@7 and not just gcc-7 <rekado>efraim: maybe. The difficulty is in keeping all users of groovy on the same version of the JDK. We cannot have a user of a lower version take jars built with a higher version. <rekado>efraim: though we probably *should* move as many packages as we can to more recent JDKs <rekado>I don’t think it’s a good idea for us to use icedtea (JDK 8) by default, actually. <efraim>I like the idea of not having too long of a bootstrap chain, but in my unexperienced eyes using openjdk9 is like using gcc-4.2. we should move up to at least openjdk11 which is still supported upstream <rekado>I’ll try setting the default JDK to something higher in the ant build system <rekado>see how far I can build an aarch64 <efraim>looks like we might need some new ant/java11 packages to replace ant/java8 packages also <efraim>'guix build groovy --with-input=default-jdk=openjdk@11 --with-input=openjdk@9=openjdk@11' <PurpleSym1>Any news on the CI, which still seems to be broken? <rekado>efraim: some packages still need jdk 8, because they use version 1.3 of the language, which is no longer supported by jdk 9. <efraim>i see. I'll see what builds and what breaks. also building to kodi with the above substitutions <civodul>ACTION books train for the Guix Days + FOSDEM <civodul>rekado: yup, i replied minutes ago :-) <rekado>civodul: the reason why I don’t like using update-spec on packages is that in the worst case we wrap hundreds of packages only to unwrap them a little later <rekado>‘guix refresh -t cran’ would have a thousand or so packages be put in an update-spec record just to be taken out of it again. <rekado>given that ‘guix refresh’ is already pretty expensive I’m wary of adding even more allocations and garbage to be collected <paul_j>Good morning Guix! Quick, and hopefully simple question: I am missing the "use-modules" for a function in my configuration: config-file. What is the best way to find where these functions live, so I can add the missing use-modules line? I am using guix system, as opposed to running Guix in a foreign distribution. <paul_j>It may be a guixrus function? I looked in the Guix programming manual and can't see a reference to it. <civodul>rekado: hmm right; i think those allocations represent a tiny fraction of the time of 'guix refresh -t cran' (say) <civodul>ideally it would be best avoided but that makes the code a bit trickier i think? <rekado>truth be told: my first instinct was indeed to just let the function return a list of <update-spec> values. <rekado>it is pretty to have a single predictable type instead of a list of two types that need to be disambiguated <rekado>on the other hand, the disambiguation is confined to the very same module and it is trivial with ‘partition’ <rekado>but it could be argued that traversing the collection to evaluate the ‘package?’ predicate on all elements is just as ugly as unconditionally wrapping and unwrapping values. <rekado>so… either approach is fine with me. They both make me equally unhappy :) <civodul>yeah i agree it's not really satisfying <civodul>i guess i tend to favor semantic clarity over performance in cases like that <civodul>let's do that so we can update things again! <civodul>alright, thank you, and apologies for the breakage <munksgaard>PurpleSym: What's the status on wip-haskell? Should I take a look? <mekeor[m]>paul_j: not sure if this is the best solution but i would ripgrep the guix tep for (def.*THING or so <paul_j>mekeor[m]: Thanks for this. I am beginning to think I am missing something. I have grep'ed the guix source and the guixrus source for the definition of config-file, but failed to find the source. The definition is in a home-emacs-configuration, with the line (init-file (config-file "emacs.el")) - I am right is understanding that "config-file" in this context is a function? I a beginning to doubt myself! <rekado>I think it would be nice if we had a few more ways to recover an unbootable system <rekado>a system that has a misconfigured system service, for example. You’re thrown into a rescue REPL but it’s quite unlikely that you’ll be able to actually fix things there and continue booting. <rekado>I have a system, for example, where nginx is configured, but activation fails because (getpw "nginx") fails <rekado>there’s no easy way for me to skip that service and continue booting into at least a basic system. <rekado>we can’t build anything for macos anyway for various reasons <rekado>e.g. no glibc port for macos, no free toolchain, etc <sughosha>Ok. I only knew that cross compiling for Windows is available. <civodul>i wonder if --add-header="X-Debbugs-Cc: ..." actually works <rekado>in icecat you can hit F12 and then override ‘display: none’ for ‘article.message header div.details’ <rekado>this shows you the details recorded for each message <rekado>it shows that Hartmut is a recipient of message #3 <rekado>I suppose that one was sent explicitly to Hartmut and the previous two messages used X-Debbugs-Cc? <rekado>FWIW in the raw first message I don’t see the header. <nckx>sughosha: That's valid, but note that the more common usage of ‘cross-compilation’ in Guix is cross-architecture, not cross-OS. I.e. using --target=aarch64-linux on x86_64. <nckx>So when we say you should try to support cross-building, we don't immediately mean you need to make it build with mingw :) <sughosha>And do you mean that the architecture for OSX is something different than these like x86_64 etc.? <mirai>nckx: about yesterday question about the libavif tool output, I did it that way because in most distributions they do split libavif into a lib-package and a tools-package and git already does something similar so I thought there's no harm on doing it <mirai>and in most cases, if you just want to link against libavif, no point in also having the tools present <mirai>rekado: if your configuration feeds invalid scheme code to shepherd you don't even get a REPL or a crash or any meaningful message, the system just seems like it's still starting up <mirai>nckx: what exactly is the @-command for foo-service-type ? <mirai>@deffn {Scheme Variable} ? @defvr {Scheme Variable} <rekado>even though I said it would be impractical to recover I actually did manage to recover (by corrupting the activation file in the store…) <rekado>beats reinstalling this system yet again <rekado>(it’s the cursed rockpro64 system that I can’t boot) <PurpleSym1>munksgaard: I bumped GHC to 9.2 and the importer to Stackage LTS 20.5, but haven’t started updating packages yet. This time I’d like to write a script to do the package upgrades, because the work is tedious and error-prone right now. Before that I’d also like to get some things in Haskell-land in shape: Packages currently don’t use the upstream-name property nor `haskell-uri`. Then I’d like to move #:cabal-revision into the origin, which would <PurpleSym1>ensure `guix build -S` gets the actual source code, including updated cabal file. The code for the latter is written, but needs testing and review. <PurpleSym1>Switching to upstream-name is also nearly finished, just needs some updates for the importer to actually use it instead of reg-exing the origin URL. <nckx>mirai: I say @defvar, but I think there's no standard (unfortunately). <nckx>sughosha: No, I don't think so. I'm not familiar with the Macintosh, but I think they are mostly standard aarch64 machines. You could run Guix System on them and they'd run software built with --target=aarch64-linux-gnu. But Guix cannot currently build software that runs on OS X, for proprietary software/toolchain reasons. Perhaps it never will. <mirai>nckx: I've looked into texinfo manual and checked guix's usage of it <nckx>Guix does support mingw (see ‘guix build --list-targets’) but it's not expected to have the same tier of support, IMO. <nckx>mirai: We use both, right? IIRC we don't have a rule. <mirai>looks like the few places that use defvar use it inconsistently <mirai>@defvr category name / @defvar name <mirai>the most common use is @defvr {Scheme Variable} nix-service-type <nckx>Isn't that equivalent to @defvar? <nckx>I'm no Texinfo wizard but know that one's an alias for the other + type of ‘variable’. <mirai>it's debatable whether there's value in displaying 'Scheme Variable' instead of 'Variable' as guix is... almost only (Guile) Scheme code <nckx>And Guix only ever deals with ‘Scheme’ variables. <mirai>but you have some gems such as @defvar {Scheme Variable} wesnothd-service-type <munksgaard>PurpleSym: I have not dug into #53729 yet, no. I have limited time at the moment (trying to finish my thesis), so I was waiting for an updated wip-haskell to work out from. Does that make sense for you, or are you waiting for my input? <nckx>I totally agree. But if you want consistency, I'm afraid you'll be disappointed. Unless you want to make the effort to standardise the status quo? <mirai>which gives you: Variable: Scheme Variable wesnothd-service-type <mirai>nckx: I could look into it though I'm no texinfo wizard as well <mirai>ACTION adds it to his laundry list <PurpleSym1>munksgaard: No, I’m not waiting for that. But I’m guessing that – if #53729 is unresolved – I will not be able to upgrade attoparsec (and maybe others?). So I’ll have to dig into it when I try do the upgrade. I cannot recommend prioritizing Guix over your thesis 🙂 <munksgaard>PurpleSym: I hope that the patch already submitted (probably zimoun's v2) is going to work as intended for you. I think itd used a different approach to fix the problem, but I don't remember how it differs. <PurpleSym1>Alright, I’ll find out 🙂 I hope you’re making good progress with your thesis! <civodul>rekado: if i check in debbugs.el, i don't see x-debbugs-cc, but i'm sure i passed it via --add-header like the manual recommends <munksgaard>PurpleSym: Thank you! I hope to have more time for Guix once I'm done :-) <gabber>howdy! (how) can i set an arbitrary envvar in my guix shell before i call a custom ./foo.sh ? trying the obvious way (`guix shell -m manifest.scm -- MYVAR=1 ./foo.sh`) i get a: "guix shell: error: MYVAR=1: command not found". <nckx>Does ‘MYVAR=1 guix shell …’ not work? <gabber>i don't think this is supposed to work in a --pure shell (sorry, forgot to mention that detail before) <nckx>If you want to run shell code in the environment, you have to start a shell, e.g. ‘guix shell -m manifest.scm -- sh -c 'MYVAR=1 ./foo.sh && echo Yay'’. <nckx>The ‘guix shell’ default is to spawn a shell if you don't pass it a command, but otherwise your command is not run in a shell. <gabber>this makes sense. would it be a useful feature to give `guix shell` a --envvar option? <nckx>But I didn't think it through. Plus, -e exists so 🤷 <gabber>:) but -- just so i understand correctly -- with -E i would have to set the variable in the spawning shell and expose it in the `guix shell` call? this seems hardly more convenient than `sh -c 'MYVAR=6 foobar'` <gabber>while we're at it: may i ask if the big red "Failing" badge in my recently submitted patch #59867 is standing in the way of a successful merge? <sughosha>Is it possible to use "(find-files)" excluding subdirectories? <sughosha>I tried that but still it is returning files recursively. <drakonis>there's a procedure that looks up just the files in the directory <sughosha>drakonis: I am trying to use "(scandir)" in `guix repl` but it is saying unbound variable. <nckx>gabber: Yes, I didn't mean that -E was a better solution than 'sh', but that -E existing is a possible counterargument to my implicit 'I don't think the guix CLI is the right place to set variables'. I don't think it's convincing one but wanted to point it out. <nckx>gabber: I doubt that it's the reason, someone would have pointed out that it was failing if so. <nckx>*That said* it's in your own interest to be proactive, see what's failing and why, and either fix it or explicitly follow up with a 'qa says this breaks x but this is why that's inaccurate'. Both will serve as reminder pings at the same time. <gabber>nckx: Ok, cool :) I am mostly puzzled by the lint warnings explanations mentioning "line 1020 is way too long". I may not interpret my patch correctly, but AFAICT it doesn't touch any line 1020 in any file -- having a more specific output (i.e. path) together with the line-length warning would be smashing <gabber>nckx: i think the whole QA frontend thing was introduced between my patch and now. i don't think i had linter problems with my patch -- i try to follow the "Submission Patches" section zealously ;) <nckx>I'm mobile and can't even find the QA lint failure, let alone conveniently run it myself. <gabber>np :) i'll patiently wait and check from time to time whether it's been merged. it's the beauty of the thing that i don't need it to be merged to mainline to reproduce it on different machines <nckx>Are you really a gabber? I like the idea of someone hacking away at Guix at 200bpm. <attila_lendvai>there are no substitutes for wine and vlc for a couple of days now <gabber>nckx: Not sure if i (or anyone) can be *a* gabber, but it's both a pun on my given name and a wink to the not-too-beloved genre of electronic music which i do -- at least occasionally -- enjoy <itd>munksgaard: Yeah, your thesis has higher priority. I can wait. :) (Can't really comment on the differences between the patches, my guile-foo is too weak for that. Just that one seems to work better for me.) <nckx>One can absolutely be a gabber. But probably only in Dutch. <nckx>I'm probably late to the pity party but we seem to have lost build history on berlin. <gabber>nckx: huh! it seems to be a valid term for people in other languages, too! so i might be one? i'll have to upgrade (or downgrade, depending on perspective) clothing and hairstyle, but i definitely won't regard it as offensive if you (and others) keep imagining myself as a true gabber, smashing my keys to fully distorted 200+ BPM techno music <__abbe__>Anyone has any idea how to install Guix in dual-boot setup with NixOS ? <__abbe__>I would like to try Guix without installing it in VM. I use ZFS for my NixOS setup. <bjc>zfs is going to be problematic with guix <civodul>__abbe__: hi! i suspect it would be hard, because both OSes want to control what goes into grub.cfg <civodul>an extra level of indirection always helps <bjc>depends on what you need <bjc>zfs on root won't work <__abbe__>I'm thinking of creating a volume for `/gnu` like similar to what i have for `/nix` <bjc>and depending on what you need mounted by login time, there are other difficulties <civodul>__abbe__: zfs is probably not well supported out of the box on Guix System, dunno <__abbe__>and another `root` volume for Guix, similar to what I've for NixOS <bjc>also, i'm not even sure if the current zfs works with the current kernel <bjc>if only someone had submitted a patch to upgrade zfs to 2.1.7 to solve that problem *cough* <__abbe__>I don't mind running older kernel. I'm already using 5.15 <__abbe__>BOOT_IMAGE=(hd5,gpt1)//kernels/cqg7lf7cvrjj5n23334w8w35d42q5bpi-linux-5.15.81-bzImage init=/nix/store/d72lypjgyz8lrx2lr7wlpzsyc8ajwdkj-nixos-system-chateau-23.05.20221228.e182da8/init elevator=none is my /proc/cmdline <__abbe__>How would I go about setting it up ? boot from Guix USB image, and then set everything up ? Or can I set it up from within NixOS ? <apteryx>managing guix-patches submission is fun using patman <KarlJoad>Does Guix have an fstrim service already available? <katco>hallo guix. should i expect guix home reconfigure to restart the user-instance of shepherd? <bjc>it should restart changed service definitions, i believe <katco>i've removed some services because reconfiguring was throwing an exception, but it never restarts the shepherd instance. i.e. it's not pointed at the new .conf <katco>haha, well done led-lightbulb_ :) <mirai>KarlJoad: last time I checked no but this can be done with a simple mcron job <KarlJoad>mirai: That was my plan. Just wanted to see if I was missing something. Perhaps I will package one up... <mirai>KarlJoad: if you do, a starting place would be the mcron service-extension <jlicht>apteryx: should guix's version of `patman` Just Work /w the guix git repo .patman? <apteryx>you still have to submit the cover letter manually to debbugs, and then add the 'Series-to: xxxxx@debbugs.gnu' patman git tag before sending the rest, but once that's done, it avoids you loosing the useful bits for later revision <apteryx>and the 'Series-version: 2' and 'Series-changes: 2\n - Modified this and that' will produce patch annotation very useful for the reviewers (and yourself) <apteryx>it'll also automatically take care to cc the right people (per etc/teams.scm) <jlicht>apteryx: hmm, on a fresh guix I get "WARNING: Unknown setting get_maintainer_script" /w guix shell patman -- patman status, but all the rest seems to work. Thanks! <podiki[m]1>is there a make command or other script for when a guix checkout complains about various .texi files having unknown references? <podiki[m]1>I find myself deleting random stuff in doc/ and restoring to work around it <apteryx>jlicht: you probably need to re-run ./configure <apteryx>that etc/teams.scm gets templated by the configure script <apteryx>and it needed an update to learn the "get-maintainer" action <apteryx>you're welcome! I hope you find it helpful too! <apteryx>there's one last detail to know: before submitting you'll probably want to strip the patman's specific git tags; it does so when producing the patches, so you can 'patman -n' (dry-run), which will leave the cleaned up patches in your tree, then you can 'git am *.patch' to apply the cleaned up patches to master before submitting <podiki[m]1>about time I make my first commit! I've checked make authenticate is successful and shows the proper (sub)key; anything else to check before a push? <vivien>Hi! I made progress on my issue’s QA yesterday. So now the derivation builds fine, but the "package changes" section reads: "Comparison unavailable". Do you know what that means? https://qa.guix.gnu.org/issue/60358 <podiki[m]1>...is cuirass not building anything? all derivations recently are red/yellow Xs <lfam>Looks like it started to fail during commits e84f17e..8e883dc <lfam>I sent a message to guix-devel <podiki[m]1>usually something is mentioned and noticed here pretty quickly, but maybe with the post-holidays it got missed <lechner>KarlJoad / Hi, did sneek tell you about my email setup? <podiki[m]1>lfam: I'll resist making a joke response to your message about New Years partying for our hardworking build farm... <lfam>They are still doing some jobs besides 'master', so it's a selective hangover <ham5urg>Did anyone installed Guix inside a remote VM? I've ordered a VPS but am not allowed to boot my own ISO, neither do I have a terminal besides SSH. All I have is a fall-back-debian-10 and the mounted harddisk under /repair. How should I try to install Guix into it? I would like to use a /etc/config.scm while inside the fall-back-debian-10. <lfam>ham5urg: The command `guix system init` can probably be used. It may get complicated in a variety of ways <ham5urg>The fall-back-debian-10 (buster) does not provide a guix.deb and its backport neither. Is there a buster-guix-repo anywhere? <nckx>Install guix with guix-install.sh, then guix system init with a system.scm that you wrote or otherwise provide. <nckx>It's not that we don't support Debian-packaged Guix but it's not required or packaged by us. <ConvolutedSquare>ham5urg: I have done it in the past. What I did was create a 2GB partition, dd the installer ISO to it, then manually added a grub entry to boot to the kernel in the ISO. <ConvolutedSquare>Might be an easier way but that's what I had to do at the time. Don't recall why exactly. <nckx>Why 2 GiB? Isn't the installer <1GiB? Did you do some clever A/B switcheroo later? <nckx>Or did you just want 2G swap anyway :-) <ConvolutedSquare>I wish I took better notes now. Might've been arbitrary. I don't recall, sorry. <lfam>Huh. Just got an error that I can't reproduce, while trying to build an updated VLC <lfam>guix build: error: /home/leo/work/guix/gnu/packages/video.scm:1940:2: package `vlc@3.0.18' has an invalid input: ("_" #<syntax-transformer pkg-config>) <lfam>It only happened once. Weird, right?? <nckx>Hardly reassuring indeed. <ham5urg>Will guix-install.sh ask into which directory it installs? Or will it shred "/" and reinstall inside the current fall-back-debian-10 ? <ham5urg>Ahh, I was hoping for something like an interactive-debootstrap-for-guix. To install into /somedir <lechner>KarlJoad / thanks! my logs were inconclusive <ham5urg>As the fall-back-debian-10 mount the "/" under "/repair" <ConvolutedSquare>Btw nckx, I think I figured out what I need to do in order to boot from the USB but mount the store from a BTRFS volume. I need to write an activation script that fixes up the grub.cfg paths(they get changed when I added a file-system that mounts to /gnu/store), mounts /gnu somewhere(to get access to the USB's /gnu/store), and copies all the kernel/initrd stuff to that new mount point via `cp -a <ConvolutedSquare>(grep /boot/grub/grub.cfg -Poe '\\/gnu\\/store\\/[^;\\s\\/]*') /.gnu/store/` <nckx>I don't know what debootstrap is, but Guix System is installed to a directory using 'guix system init'. <ham5urg>Ahh, nckx. First step to install guix into debian and then do 'guix system init' <nckx>ham5urg: Sounds like a standard Guix installation. Because you're not in the installer OS, you won't have cow-store, so skip that step in the docs, and of course substitute /repair for /mnt. <nckx>Sorry, very slow typing. <nckx>I menat 'Standard Guix System installation'. <lfam>Especially the section on Manual Installation <nckx>ConvolutedSquare: I'll read that with more attontion later but thanks! Intereshing to follow along. <ConvolutedSquare>nckx: mind you, this is only necessary because the device's firmware only seems to allow access to the first two drives during boot. Why, I don't know. So i doubt this will be very useful knowledge :P <nckx>ConvolutedSquare: I was determined not to lazily blame your firmware, and yet here we are. What a terrible bug. Congrots for finding it. Ttyl when I don't sound drunk. <podiki[m]1>committers: what are your recommendations for workflows for patches (emacs/magit especially)? browse debbugs in emacs, something to apply the patch (sign and sign-off)? <lfam>I could share my workflow but it's very different from what you asked. It's based on mutt <lfam>Can't say I recommend it <lechner>does anyone use emacs to browse debbugs? <podiki[m]1>I have, briefly, but no expert; used it to comment on bugs that were not ones I submitted (but it didn't CC the bug number in the email by default, just the submitter I think) <podiki[m]1>lfam: mutt...that takes me back to being young, the age before webmail (of course now I use emacs for email) <lfam>Heh, I started with mutt long after webmail <podiki[m]1>while I got you here, is there a timeline for making linux-libre 6.1 the default? in a week or two? <lfam>Hm, I hadn't thought of a timeline yet. Probably within 7 days <lfam>And before that, make sure that ci.guix.gnu.org is building the master branch <podiki[m]1>looks like no x86_64 workers are doing anything, but some (all?) of the others are working <podiki[m]1>maybe the reconfigure of berlin will unstick the workers too.... <phodina[m]1>What's the proper way to add to u-boot additional configs? As I add some it fails at configure stage by throwing `Mismatching configurations in .config and configs/quartz64-a-rk3566_defconfig (("CONFIG_LOG_MAX_LEVEL" (#f "7")))` <lfam>podiki[m]: Yeah... maybe. I'm afraid to touch it <nckx>lechner: Would be nice if the bot could reply by nick, not account name (~). <podiki[m]1>also: belated happy 2023 guix! (I've been away a bit but slowly getting back now) <phodina[m]1><phodina[m]1> "What's the proper way to add..." <- I've come up with workaround that phase after `configure` modifies the file `.config`. It's not the best way but works now. <davidl>Hi. I need to use the btusb linux firmware module. How can I do that? <davidl>I can't just add (firmware (cons* btusb %base-firmware)) to my config.scm and reconfigure cuz it won't find btusb. <davidl>the %base-firmware doesn't contain much. <sughosha>Hi there, is it recommended to use texinfo format for comments also, like ";; @code{wdl}"? <davidl>podiki[m]1: it is a linux kernel module btusb.c that contains firmware for usb bluetooth devices, such as asus usb-bt500 <podiki[m]1>does modinfo btusb show you anything (I'm not on linux-libre)? <podiki[m]1>davidl: so I think you are just wanting to load the kernel module, not a firmware <jackhill>hmmm, gnome-control-center is segfaulting for me with sway. `guix gc --verify=contents,repair` doesn't turn up any problems. Any ideas? <bjc>i have one of those asus bt500 adapters and just tried it; the kernel module loads automatically, but: “Jan 3 14:36:06 localhost vmunix: [369475.454005] hci0: Missing Free firmware (non-Free firmware loading is disabled) <bjc>so it looks like it cans't be used in vanilla guix <davidl>podiki[m]1: hmm. Maybe. I tried using info from that link, but it's not straight foward, I wonder if I need to define a btusb-config file. <davidl>podiki[m]1: supposedly it works on trisquel 10 nadia according to h-node.org <bjc>maybe somewhere. i've never checked. i've only ever used this adapter with manjaro before <davidl>podiki[m]1: and yes, modinfo btusb does show me that it exists. <davidl>blueman-adapters 20.41.50 ERROR Adapter:63 __init__ : No adapter(s) found <davidl>still after restarting bluetooth service etc. <bjc>the only drivers i can find for the adapter are blobbed <podiki[m]1>alas I've never used bluetooth (on a computer)....always seemed confusing to me too <davidl>podiki[m]1: so, you are probably on the right track - the device driver exists as free software, but maybe no free firmware so the device won't be found. also, dmesg says: Bluetooth: hci0: RTL: firmware file /*(DEBLOBBED)*/ not found <davidl>(s/won't be found/won't be accessible?/) <podiki[m]1>if indeed nonfree firmware is required, linux-libre won't load it <lfam>thinkpenguin.com does sell a bluetooth dongle, so presumably there is a free implementation <jackhill>Anyone else notice GTK4 apps not taking up the full height on sway portrait displays? Is it a GTK bug or something in our packaging? <davidl>I don't fully understand the difference between firmware and device drivers :-( <podiki[m]1>so that might be similar to most new-ish GPUs in that there may be a free driver (amd in mesa) but effectively can't do most anything without some firmware blob <jackhill>narrow windows on landscape displays seem ok <lfam>davidl: There's a difference, it's a bit of a line in the sand in 2021. The upshot is that a lot of things with "free drivers" won't work on linux-libre <davidl>podiki[m]1: Im also considering switching to a normal linux kernel to not having to deal with this, and so I get microcode updates etc. <lfam>I doubt the thinkpenguin dongle requires non-free firmware, since it specifically says that it does not <bjc>the thinkpenguin adapter it bt4, the asus bt500 it bt5 <sner>davidl: drivers are cpu code, firmware is device microcontroller code <podiki[m]1>davidl: well if you go that route you'll have to ask about that kernel and all that in that channel's irc ;) <davidl>podiki[m]1: I promise :) I had hoped to avoid it but Im not sure what the best choice is anymore. <podiki[m]1>not to get into the whole thing here, but the line is rather fuzzy as many devices (like a cpu) will have microcode when you buy them already, but you can't update it/change it on linux-libre (I know this is a sore spot for many with FSDG) <lechner>Hi, how can i find out if this BT dongle requires a blob, please? 0a12:0001 Cambridge Silicon Radio, Ltd Bluetooth Dongle <podiki[m]1>davidl: understandable, and sadly the state of affairs for certain hardware these days <bjc>when it says "does it work with free software?" does that mean blob-free? because btusb will happily load a blob if it can find one <bjc>according to the same website, the bt500 also works with "free software" via btusb <podiki[m]1>the "tested with kernel libre" I assume means linux-libre which will have no firmware <bjc>hmm. maybe there's a guix option then, but it's just not packaged up, or needs some config? <bjc>might be worth looking into what trisquel is doing <nckx>bjc: <when it says> It doesn't mean anything, sadly; not everyone knows what firmware is or that Linux® includes it by default. H-node has mistakes. <efraim>what do we have for (assoc-ref %build-inputs "rav1e") when in a gexp? <nckx>search-input-file? Why can't you leav the gexp to use this-package-input? <lechner>sner / thanks! I was lazy. the -r option is missing from the manual page <the_tubular>Is there a new way to lock a specific package to a version ? <nckx>Guix has man pages? Nuts. <nckx>Ah, they are auto-generated from --help, and -r is missig from --help. Now it's a real bug. :) <sner>nckx: in the infodoc it seems the original `-r` flag was deprecated, but left as an alias for `--serializer=nar`. Might be worth noting that its an alias <leg7[m]>Turns out I just needed to use luks1 <leg7[m]>Config looks like this except that the mapped-devices source is the luksUUID shown in the boot error and not the template block device "dev/sda2" <leg7[m]>From what I understand this error is the guix initramfs failing to access my system because the disk is encrypted <the_tubular>Hey civodul, i was looking at a330bfdf5b2f27c34b5ece82bb355553af110cfe does that mean I don't have to inherit a package to lock it to a specific version ? <leg7[m]>Ok I reinstalled guix and the initrd bug fixed itself <Kolev>Should I use wlsunset or Gammastep? <nckx>sner: So my plan was to (a) document the TYPEs, like it already documents ALGORITHMs above and (b) maybe add a note about ‘-S nar’ since it's such a common case (c) note ‘-r’ in that note, because ‘-rx’ is so ingrained, but not in the actual list of options. <nckx>But maybe people really want ‘-r’ to die, I don't know? <apteryx>the_tubular: to lock a package to a specific version, you could read into using inferiors <apteryx>you can peg packages to their exact guix revision <the_tubular>I only need 1-2 package from the main channel locked <the_tubular>I was just wondering if that new commit was something else I could use <civodul>no, but it'll be the basis of a new --with-version option, similar to --with-latest <bjc>is it me, or are the hunspell packages a bit of a mess? <bjc>they're scattered across, at least, libreoffice, aspell, and hunspell modules, the last of which only contains the italian dictionary <bjc>also, it can't lookup dictionaries installed via guix home, from what i can tell <bjc>i guess what i'm asking is: is it worth my time to try and consolidate all this under hunspell.scm? <rekado>I also think they are a bit messy <bjc>alright, i'll get to work on it, then =) <rekado>sughosha: would you be annoyed if I made some comments about cosmetics / style? <rekado>when you use a seemingly arbitrary commit it would be good to add a comment (e.g. to state that there has been no official release yet) <rekado>for the snippet I recommend (with-directory-excursion "WDL" (for-each delete-file-recursively '(…))) instead of using string-append <rekado>I’d also break the line after “(snippet” <rekado>the values in (arguments …) are pretty shifted to the right; better add a line break after “(list” and again after “#:phases” <rekado>the check phase should respect the #:tests? argument, so that the --without-tests package transformation can be used. <rekado>but perhaps you could get rid of the custom phase all together with #:test-target "test" instead. <rekado>in the 'install phase you do mkdir-p first and then copy-file <rekado>maybe you could use install-file instead, which creates directories before copying. <sughosha>I tried #:test-target "test", but the problem is I have to change directory. <rekado>it’s a pity we can only give it a target, not arbitrary make flags that only affect the test phase <rekado>another option (hardly better) would be to add a pre-check phase that does chdir. But that’s stateful, which makes it uglier. <sughosha>I wish there would have been something like #:test-flags <rekado>the description’s first sentence is a little … awkward <rekado>“WDL is a C++ library that is designed to be easily used by other code.” <sughosha>I know but that's what I could find on the website 😂️ <sughosha>"WDL is a modestly reusable C++ library with the following features:..." <rekado>one last comment about the patch: would it be possible to set LFLAGS or LINKEXTRA instead of patching the Makefile? <rekado>you also remove one object file, but the patch summary is too short for me to guess as to why <sughosha>I tried but no. LINKEXTRA is applied only for swell, but not for eel2. <sughosha>For eel2 I could not find any other way than a patch. <rekado>perhaps you could add a short description to the top of the patch? This would also make it easier for future maintainers to see if it’s still needed or if it’s been upstreamed, etc <sughosha>There is also a bug with jnetlib subproject, it's Makefile lists an invalid object. <sughosha>Would you recommend to have different patch for bug and for build arguments (LINKEXTRA etc.)? <sughosha>rekado: install-file takes second argument as a folder, not the target file. The problem I faced was that I had to have the same hierarchial order of the source in the target. <sughosha>find-file returns relative file path, so copy-file could take it and then install it relatively. <ham5urg>Is it possible to let bootloader in config.scm unconfigured? I try to install into a remote VPS which will use the hosts kernel. Maybe it's a container, not a VPS. Idk. <bjc>--no-bootloader for 'init', do not install a bootloader <ham5urg>I tried 'guix system init --no-bootloader config.scm /repair' but get 'missing field initializers (bootloader)' <bjc>you should be able to put in a dummy value there, and it'll get ignored (or error out if you forget the --no-bootloader switch) 