IRC channel logs

2020-05-12.log

back to list of logs

***skangas_ is now known as skangas
<gagbo>sneek: seen davet ?
<sneek>daviid was in #guile 4 days ago, saying: unknown_lamer: guile.m4 has been patched a couple of times since, I would grab the latest and use it to replace the one you keep in yur project.
<gagbo>sneek: later tell davet hi, I was wondering how you'd like to see contributions to guile-sdl2 ? I'd like to access the font-metrics functions from SDL_ttf and I don't know how hard it would be for a guile beginner to make, and I don't know how to submit a patch if I ever manage to do it myself
<sneek>Okay.
<gagbo>oh, didn't work at all as expected
<gagbo>I guess I'll have to lurk the old way
<unknown_lamer>dsmith-work: does sneek have a tell function?
<gagbo>it does, I've seen it work, I just didn't read that my query for davet wasn't met with davet, but daviid ; so I'm not sure that my request will ever land
<gagbo>I guess I'll have to send emails, but I don't know how to build properly guile-sdl2 from source now, so that won't help me patch it to get what I'd need. Looks like I'll have to sink a lot more time than I imagined
<lfam>gagbo: Are you looking for the maintainer of guile-sdl2?
<gagbo>I was
<gagbo>I still am, I guess
<lfam>I don't know how much time they have for this right now, but that person is davexunit
<lfam>We have a package for guile-sdl2 in Guix. It seems to build, so you could take a look at that to get started, or even just do your work in Guix
<gagbo>Oh yeah, I'm actually using the packaged version
<lfam> https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/sdl.scm#n620
<gagbo>That's why I know how I can use it, but not really how I can patch it and test my features
<lfam>Oh, alright
<lfam>If you want, I can help you get started over in #guix
<gagbo>I can try quickly
<dsmith>unknown_lamer: I think I've diabled "tell" because it doesn't do what most people expect. "later tell" is the usual thing to use.
<dsmith>unknown_lamer: "tell" just causes the bot to send the message immediately. If the nick is in the channel or not. Or away or not.
<unknown_lamer>useful heh
<dsmith>The only real use I can think of, is if the target nick is /ignore'ing the sender.
<dsmith>"later tell" waits for the nick to message the channel before sending the messages.
<dsmith>Might be nice to change that for joins instead.
<lfam>I tend to join and leave randomly when my laptop sleeps and wakes
<dsmith>unknown_lamer: For that, the bot would need to keep track of who is in the channel, nick changes, etc.
<dsmith>Not currently doing that.
<lfam>I like that sneek messages when I speak, becauase it's a sign that I'm ready to participate
<unknown_lamer>makes sense
<dsmith>Yep
<lfam>sneek: botsnack
<sneek>:)
<lfam>Good bot
*sneek wags
<lfam>!!!
<lfam>Such a good bot
<unknown_lamer>need to add a hook that runs after the bot joins a channel *and* the userlist is initialized
<unknown_lamer>or I wonder if there's any reason not to move hooks/join to after that
<unknown_lamer>not sure what use there is in running a hook right when you join before any info on the channel is available
<dsmith>I can hook on joins. And leaves. And nick changes.
<unknown_lamer>right, but when you join the channel you need to query the roster to get the initial state?
<dsmith>IF I hook on raw maessages, I can use the 532 code to populate the nick list.
<dsmith>The server sends those.
<dsmith>Not sure if it's because of a query by bobot code, or if it just does that automatically.
<unknown_lamer>bot does a WHO after joining
<dsmith>ok
<unknown_lamer>but I think hooking on raw messages is ... hacky
<unknown_lamer>I try to think of reasons against moving the hook to after userlist initialization, but right now i think moving it is the right thing
<dsmith>YEah, I don't really like it. Especially since the bobot code is already handling that.
<dsmith>What might be better, is if the scheme code could just query the C++ userlet. Instead of bulk loading the whole thing.
<dsmith>gagbo: Yeah, original sarahbot code uses a "metaphone" encoding for nicks. I think it's so variations in spelling don't matter.
<dsmith> https://en.wikipedia.org/wiki/Metaphone
<dsmith>----
<tohoyn>sneek, seen daviid
<sneek>Welcome back tohoyn, you have 1 message!
<sneek>tohoyn, str1ngs says: It does not work with guile 3.0 as far as I've tested.
<sneek>daviid was in #guile 4 days ago, saying: unknown_lamer: guile.m4 has been patched a couple of times since, I would grab the latest and use it to replace the one you keep in yur project.
***apteryx_ is now known as apteryx
<civodul>"invalidate-compilation-cache is deprecated; recompile your modules"
<civodul>oh i see
<wingo>civodul: yay
<wingo>i finally managed to trim cps out of the bootstrap module graph
<wingo>in git from a couple hours ago
<wingo>so bootstrap should be feasible now :P maybe 20 minutes -j1
<civodul>wingo: yay, awesome!
<civodul>in other news, reepca found that the revealed count of file ports wasn't initialized, oops!
<civodul>i'm writing tests for that
<wingo>!
<wingo>that's exciting :P
<dsmith>Ooo
<dsmith>I wonder if that was causing the "random" startup crashes I was seeing on arm.
<dsmith>And it would go away if I used a pipe for stdin.
<wingo>!
<dsmith>OR maybe it was stdout? Was just repeatedly starting guile in a loop, seeing if it segfaults.
<dsmith>Redirecting to /dev/null to make things quieter stopped the crashes.
<dsmith-work>Tuesday Greetings, Guilers
<wingo>guix is building guile 3.0.2 and i am already irritated that it is taking so long :P
<wingo>not 100% fixed in 3.0.3 but at least the "long pole" of compiling psyntax is only 3m instead of 12 or whatever it is now
<civodul>heheh :-)
<civodul>wingo: it's building it from source, why is that?
<wingo>what do you mean, why is building psyntax faster in master?
<wingo>or why is guix building from source
<wingo>i guess because there was some update to mes or something
<wingo>it rebuilt everything. i passed --substitute-urls=https://ci.guix.gnu.org/ but no dice
<wingo>civodul: ^
<rekado>does /etc/guix/acl exist?
<civodul>yeah maybe the server is not authorized: https://guix.gnu.org/manual/en/html_node/Substitute-Server-Authorization.html
<civodul>you should definitely get binaries for Guile
<wingo>hum
<chrislck>wish to ask devs here... "too many heap sections" means, I believe, a gc issue. Is there a magic flag we can set to increase the heap limit?
<wingo>chrislck: i have never run into that, myself; perhaps look at libgc documentation
<wingo>weird, it still wants to build guile from source; no idea why
<civodul>you could "rm -rf /var/guix/substitute/cache" for good measure, though it shouldn't make any difference
<wingo>yeah no difference
<wingo>i haven't rebuilt in a while, surely not since 3.0.2
<wingo>so maybe it's just missing that substitute? dunno. weird
<chrislck>wingo: thx for pointer. will do. some gnc reports are crashing on 32-bit windows.
<wingo>that could indicate a memory leak fwiw
<rekado>I see a whole bunch of guile-3.0.2*.nar* files in both the gzip and lzip caches on ci.guix.gnu.org
<wingo>ok
<wingo>i admit, on this ubuntu machine i don't understand everything. but it seems to want to build from source today. standard "guix pull"
<wingo>acl in place, no warnings
<rekado>strange!
<civodul>wingo: what does "guix build guile --no-grafts -d" return?
<civodul>for me it's /gnu/store/bl4pvflvbw0lyk0y05f2md94kgyrc6an-guile-3.0.2.drv (x86_64)
<wingo> /gnu/store/cgk36zvqbd9199yhi0ywfy2lmm20zxdz-guile-2.2.7.drv
<wingo>but this is in a guix pull
<wingo>so i don't know if the "new" profile is in place yet
<civodul>so that's prior to the recent "core-updates" merge
<civodul>ah you mean "guix pull" is building this one?
<wingo>no
<wingo>the drv above is what "guix build guile ..." prints
<civodul>ok, and what does "guix describe" return?
<wingo>but guix pull is building guile 3.0.2, before showing me what it's going to do
<civodul>hmm, ok
<wingo>Generation 15 May 05 2020 11:49:10 (current)
<wingo> guix 0e68092
<wingo> repository URL: https://git.savannah.gnu.org/git/guix.git
<wingo> branch: master
<wingo> commit: 0e680920b9e54d7c8a901b1c9cf02ce4468f44ed
<wingo>probably i am an idiot somehow but it's partly because i am an idiot that i switched over to the "guix pull" workflow, rather than running from my dev dirs :P
<civodul>well maybe there's an actual problem too, that's more likely than you being an idiot :-)
*civodul tries guix time-machine --commit=0e680920b9e54d7c8a901b1c9cf02ce4468f44ed -- time-machine -- describe
<civodul>first time i tried such a command, i find it interesting
<civodul>it worked, but that doesn't tell much
<janneke>i have found it to be extremely useful
<civodul>perhaps i happen to have the stuff in store
<civodul>hey janneke
<civodul>the double time-machine?
<janneke>ah no, i missed that
<janneke>the single one :-)
<civodul>ah, see? :-)
<civodul>"guix time-machine --commit=0e680920b9e54d7c8a901b1c9cf02ce4468f44ed -- weather guile-next@3" says 100%
<wingo>guile-next is fine i think
<wingo>well
<wingo>build-started /gnu/store/x32cnfkd50fnxs10xp1jdn24h7ai2gxr-guile-3.0.2.drv - x86_64-linux /var/log/guix/drvs/x3//2cnfkd50fnxs10xp1jdn24h7ai2gxr-guile-3.0.2.drv.bz2 1726316
<wingo>that is what it is building
<wingo>for some reason it is showing the build log, interleaved with @ build-log
<wingo>just a normal "guix pull"
<civodul>ah got it, it's: guix build -e '(@@ (gnu packages commencement) guile-final)' -d
<wingo>wingo@milano:~$ guix build -e '(@@ (gnu packages commencement) guile-final)' -d
<wingo> /gnu/store/y88s25rjwdic7x2ijfqn2mjszias0kvs-guile-2.2.6.drv
<civodul>yes, but the guile-3.0.2.drv above is the new "guile-final"
<wingo>ah
<wingo>ok :)
<civodul>substitute here: https://ci.guix.gnu.org/0m0vd873jp61lcm4xa3ljdgx381qa782.narinfo
<wingo>it built gcc and mes and everything to get there, fwiw
<civodul>that means you really got zero substitutes, right?
<civodul>then perhaps the key in /etc/guix/acl is the wrong one?
<wingo>i guess? it did make a number of lines like "getting substitutes from https://ci.guix.gnu.org/ 100%"
<wingo>and no warning about bad acl things
<civodul>yes, but the authorization check happens after
<civodul>it's not "bad", it's just unauthorized
<civodul>see https://git.savannah.gnu.org/cgit/guix.git/tree/etc/substitutes/berlin.guixsd.org.pub
<wingo>that's the one in /etc/guix/acl
<wingo>the guix daemon that is running is the one from /var/guix/profiles/per-user/wingo/current-guix/bin/guix-daemon, fwiw
<civodul>ok, and do you have /var/guix/substitute/cache/4refhwxbjmeua2kwg2nmzhv4dg4d3dorpjefq7kiciw2pfhaf26a/0m0vd873jp61lcm4xa3ljdgx381qa782 ?
<civodul>wingo: in other news, a .go file compiled with master -O0 makes 3.0.2 segfault
<wingo>civodul: nope
<wingo>i don't have that substitute
<wingo>civodul: yes i don't doubt it, a .go file from master uses new instructions from master
<civodul>right, but i thought there was this version cookie
<civodul>but maybe it's not incremented yet?
<wingo>ah, yes indeed. not yet incremented
<civodul>ok
<wingo>i was trying to build guile on this machine but guix got in the way ;)
<civodul>yeah it's terrible, i don't understand what's going on
<wingo>so why wouldn't i have the substitute?
<wingo>there were like 50 lines of "getting substitutes" and they all went 100% immediately
<wingo>my network is really slow btw
<wingo>so that's somewhat unusual
<civodul>what does "sudo grep -r 0m0vd873jp61lcm4xa3ljdgx381qa782 /var/guix/substitute/cache/" return?
<wingo>could the substitutes server be returning truncated output or something?
<wingo>civodul: nothing
<civodul>nothing is very little
<civodul>it's as if substitutes were turned off
<wingo>yes it seems that way. maybe it is a permissions issue somehow
<civodul>yeah is /var/guix/substitute writable by the daemon? is it empty?
<wingo>so
<wingo>having removed the cache
<wingo>there is only one dir now
<wingo> /var/guix/substitute/cache/x2wcz6gz3evwlqcrz3fqstmezkfcfnpfb5kfyxbz7kjikc7upkiq
<wingo>which has a bunch of 99-byte files
<wingo>whose contents are all (narinfo (version 2) (cache-uri "https://ci.guix.gnu.org/") (date 1589292026) (ttl 600) (value #f))
<wingo>i.e. i removed the cache then did a git pull.
<wingo>*guix pull
<civodul>(value #f) means the substitute was not found
<civodul>ttl 600 means it'll try again in 5mn
<civodul>is https://ci.guix.gnu.org/0m0vd873jp61lcm4xa3ljdgx381qa782.narinfo accessible from that machine?
<wingo>yes, it does return information
<wingo>at least via wget
<dsmith-work>git master: real: 60m12.574s user: 140m37.284s
<dsmith-work>git v3.0.2: real: 77m18.208s user: 189m15.317s
<dsmith-work>tar v3.0.1: real: 9m52.616s user: 31m53.711s
<dsmith-work>Uh, s/3.0.1/3.0.2/
<civodul>for python-xyz.scm it seems to go from 30s at -O1 to 39s at -O0, weird
<wingo>civodul: that sounds like you are not running the right guile
<wingo>or, there's some O(n^2) pathology somewhere; but i suspect not running the right guile
<wingo>dsmith-work: what master?
<civodul>wingo: i'll investigate and report back later
<wingo>an unsatisfying day for both of us then :P
<dsmith-work>v3.0.2-76-g3385f6e08
<dsmith-work>"Update NEWS for baseline compiler"
<wingo>interesting
<wingo>is this on your ARM machine or an x86_64 machine, dsmith-work ?
<wingo>and "tar" you mean with prebuilt/ i guess
<dsmith-work>x86_64. 4 cores. make -j5 (but is in a vm under w10)
<dsmith-work>Yes, tar is from a tarball build
<dsmith-work>Using the prebuild .go files
<dsmith-work>So. Better. But not quite as awsome as hoped.
<wingo>well, it's tricky. the baseline compiler produces slower code so even the cps compiler as compiled by baseline will be slower, until it is recompiled by cps
<dsmith-work>YEah, I was wondering about that.
<wingo>i am not sure what is the fastest way to do things
<wingo>even the evaluator compiled by baseline will be slower
<wingo>until recompiled
<dsmith-work>Commit the prebuilt .go files to git... ;^}
<wingo>haha ;0
<dsmith-work>That's basically what ikarus did.
<dsmith-work>But was intel only.
<wingo>i think the baseline compiler will be significantly better for guix tho
<wingo>since it doesn't need to bootstrap
<wingo>i see 5x-15x speedups
<lloda>>how do I force guile-snarf-docs to be created from guile-snarf-docs.in
<lloda>ah there it is nvm
<daviid>lloda: tha's done at configure time I guess
<sneek>daviid, you have 1 message!
<sneek>daviid, gagbo says: hi, I was wondering how you'd like to see contributions to guile-sdl2 ? I'd like to access the font-metrics functions from SDL_ttf and I don't know how hard it would be for a guile beginner to make, and I don't know how to submit a patch if I ever manage to do it myself
<lloda>yah thx daviid i've managed to make generate-docs
<daviid>gagbo: the author of guile-sdl2 is davexunit
<lloda>cyclopsian's hint did it
<daviid>:)
<dsmith-work>daviid: The bot picked the wrong nick. :(
<daviid>tohoyn: g-golf 'as is' won't work with guile 3.x, because it extensively 'expand/extend' some of its modules dynamiclly, which is the very essence of introspection ... and guile 3.0 has default compilation settings that expect modules to be immutable, also, 3.0 changed the 'meaning/applicability' of @ and or @@ ... - I won't work on this until I find it's more-less complete (no specific date, dont ask ...) - so plese use 2.2.x for the
<daviid>time being ... I also have to prepare guile-cv for that same maters ...
<daviid>dsmith-work: ok, just wanted to let gagbo know ...
<tohoyn>daviid: so we won't get g-golf for guile 3.0 at all?
<daviid>tohoyn: that's not wht I said, is it?
<tohoyn>daviid: I understood it wrong
<tohoyn>daviid: is guile 3.0 missing some features needed for g-golf?
<daviid>tohoyn: please re-read the above
<daviid>carefully
<tohoyn>daviid: so do you need @ and @@ to be more complete?
<daviid>tohoyn: I need to prpare g-golf so it works ith gule 3.x, that requires some work that i'm not going to do now, but after it is complete and after i port guile-cv to 3.x as well ...
<gagbo>daviid: oh yeah no worries, I should have added that I was told later about the correct nick. I just sent a patch so he'll deal with that when he can (I forgot it was still easy to use gmail to git send-email)
***slyfox_ is now known as slyfox
<andswitch>how do I pass a keyword argument?
<dsmith-work>andswitch: https://www.gnu.org/software/guile/manual/html_node/lambda_002a-and-define_002a.html
<dsmith-work>andswitch: "can be called as (sir-yes-sir #:action 'jump), (sir-yes-sir #:how-high 13), (sir-yes-sir #:action 'lay-down #:how-high 0), or just (sir-yes-sir)"
<andswitch>ah ok :)
<andswitch>I saw that page but missed it because i was looking at the code-blocks I guess
<andswitch>th
<andswitch>thx
<dsmith-work>np
<dustyweb>hi!
<catonano>dustyweb: hi !
<dustyweb>heya catonano, how are you
<catonano>dustyweb: I could e better but I could be worse ! 😁️
<dustyweb>:)
<wingo>hum the baseline compiler is still taking forever on assembler.scm
<wingo>i wonder why that is
<andswitch>is the one-page HTML reference manual for guile 2.2 still available somewhere? I can only find a link for the 3.0 version on the homepage
<catonano>andswitch: you can find links for previous versions at this page: https://www.gnu.org/software/guile/learn/
<andswitch>catonano: yes I saw it but I can't find the link to the single-page html version of the reference manual. Only the multi-page one is there