00:46 Altai-man_ left 01:16 ZzZombo_ joined 01:19 ZzZombo_ is now known as ZzZombo 02:39 Xliff joined 06:05 jmerelo joined 06:23 Geth left 07:42 jmerelo left 08:11 Xliff left 09:10 sena_kun joined 09:30 Altai-man_ joined 09:32 sena_kun left 09:42 donaldh joined
El_Che patrickb: set-env.sh is problematic. It set relative paths in the PATH: export PATH=./../bin:./../share/perl6/site/bin:... 10:04
tellable6 El_Che, I'll pass your message to patrickb
10:47 tyil left 10:48 tyilanmenyn joined 11:31 sena_kun joined 11:33 Altai-man_ left 11:59 ggoebel left
El_Che AlexDaniel: flappers on i386: travis-ci.org/nxadm/rakudo-pkg/jobs/609193309 12:36
12:36 patrickb joined
AlexDaniel gah, that performance thing again 12:36
El_Che it's not head
but the latest release 12:37
AlexDaniel: reading relocatable tar.gz's for the release
AlexDaniel well, it doesn't matter, the test was flappy anyway… 12:38
El_Che it's a rather painful flipflop. Somethings I need 10 i386 to get packages 12:39
12:44 patrickb left
rba El_Che: Are the docker images built with alpine new and rakudo v2019.07.1-03 around somewhere? 12:56
El_Che rba: but it's no more than adding a 2 line Dockerfile to a regular Alpine image 13:01
rba: curl -sSL github.com/nxadm/rakudo-pkg/releas...x86_64.apk
apk add --allow-untrusted *.apk
rba: I haven't come around to create alpine repo's like for rpm and deb, but will do in the future 13:02
not straightforward though (I host on bintray)
I am also considering creating images for all the distributions I support, but for now working in relocatable archives 13:03
rba El_Che: Sure. I will go with the two lines for now. 13:10
13:22 lucasb joined 13:30 Altai-man_ joined 13:33 sena_kun left 13:34 Xliff joined
Xliff o/ 13:34
El_Che AlexDaniel: are t/09-moar/Line_Break__LineBreak.t and /09-moar/General_Category__extracted-DerivedGeneralCategory.t als related to performance (next to t/08-performance/99-misc.t; these keep failing on relocatable i386) 13:47
jnthn Are they failing or just have passing todo?
13:54 tyilanmenyn is now known as tyil
El_Che jnthn: travis-ci.org/nxadm/rakudo-pkg/jobs/609218558 13:54
jnthn Yes, "TODO passed" 13:55
El_Che so only the misc test is the one holding the pass? 13:56
AlexDaniel yep 13:57
El_Che ok, nothing new then 14:00
sorry for the noise
14:04 Xliff left
AlexDaniel El_Che: it's not noise, that damn test needs to be fixed 14:06
El_Che looking at Windows travis builds 14:23
it... interesting
running bash commands to do stuff
15:30 sena_kun joined 15:32 Altai-man_ left
El_Che OMG, building stuff (Rakudo) on Windows is concentrating pain. Fighting now with perl 5 to install the needed modules. 15:38
I will give managers that expect me to run Windows at work instructions to build rakudo :) 15:39
that will teach them :)
moritz El_Che: back in the days when I was a windows-only user, I remember trying to compile some stuff and it was always a HUGE pain. Even as a pretty inexperienced Linux user a year or two later, compiling stuff from source was just way easier 15:44
it's kinda sad that there hasn't really been any progress in that area lately :/ 15:45
(this was around y2k, iirc)
El_Che wel, in travis they use gitshell, so it's less painful 15:46
but still painful enough
I don't think depending on perl 5 does rakudo on windows any favours 15:50
shoot me: "Can't load 'C:/Strawberry/perl/lib/auto/Cwd/Cwd.xs.dll' for module Cwd: No such file or directory at C:/Strawberry/perl/lib/XSLoader.pm line 93." 15:56
(the file is there) 15:58
it gets better: Cwd.c: loadable library and perl binaries are mismatched (got handshake key 00000006000005b0, needed 0000000000000000) 16:08
ok, enough windows for 1 day :)
16:10 Xliff joined
lizmat hugs El_Che 16:15
El_Che on the flipside, I think I am ready to add relocatable tar.gz for Linux to rakudo-pkg (and it opens the door for Windows and macos) 16:19
16:38 lizmat_ joined 16:41 lizmat left, lizmat_ is now known as lizmat 17:07 patrickb joined
patrickb El_Che: I successfully managed to create a Windows relocatable build with those instructions. It would be interesting to find out why it ended up being such a hassle on your machine. Did you start off with a clean Windows VM as described, or did you use a preexisting Windows? 17:15
El_Che: As long as I can't reproduce the relative path in set-env.sh problem it will be difficult for me to do anything about it. Could you have a look yourself or try to find out what circumstances are needed to make it fail? 17:16
El_Che patrickb: windows == travis
patrickb Travis now has Windows support? 17:17
Your trying to do it on travis. I see. That explains the different behavior. :-)
El_Che yes, beta, but it looks good enough 17:18
patrickb: trying to reproduce set set-env thing. unpacked relocable build in my ~/Download, mounted as a volume to a debian:10 docker container 17:21
cd /mnt/rakudo-2019.07.1/bin 17:22
. ./set-env.sh
ERROR: 'bash' does not exist.
that's weird
patrickb is the second . in that command on purpose?
El_Che yes 17:23
without it's the same result
patrickb I'll soon have to leave again for another three hours or so... 17:25
Is it possible this fails only on busybox? 17:27
El_Che I started it with bash 17:28
as cmd
# echo $BASH
17:30 Altai-man_ joined
patrickb El_Che: I also tried it in a clean debian:10 container. Still no relative paths. 17:32
have to leave...
17:32 patrickb left, sena_kun left
El_Che patrickb: gist.github.com/nxadm/772b514ea812...8f9671dbbf 17:34
tellable6 El_Che, I'll pass your message to patrickb
lizmat AlexDaniel: so why has github.com/perl6 not been renamed to github.com/raku yet? 17:45
have I missed something ?
or do we need to coordinate that better to avoid breakage?
El_Che lizmat: iirc he thought it was easier moving he repos one by one 17:46
lizmat ah, ok
Kaiepi is there documentation somewhere on how the compiler handles signatures? 18:09
there's a lot of code i'd need to read if there isn't
vrurg Kaiepi: last year I failed to find any. 18:26
Kaiepi agh 18:29
i was hoping there would be because the slang for a typeclass library i'm working on needs to work with how signatures are represented at that point 18:30
vrurg Kaiepi: If you have a signature object you can just see if it ACCEPTS a capture. Or you need something lower-level? 18:32
18:32 lucasb left
Kaiepi i think i might need something lower level than that 18:33
i want to disallow parameters that are optional in any way with typeclasses
vrurg I was only tangentially related to this when worked on a bug last year. But most of the related code I found in 6.c BOOTSTRAP. 18:40
Maybe find_best_dispatchee on Routine would be a good example of manipulations with signature and parameters. 18:43
robertle does anyone know what the story behind the libtommath is that we use for moar? it seems to be a forked version of the upstream and quite behind, yet not have significant changes from us... 19:12
is there a plan to catch up to a current version?
19:30 sena_kun joined, sena_kun left 19:32 Altai-man_ left
moritz has libtommath has a release yet? 19:44
afaict the problem was that some distros like debian don't like us fat-packaging libraries that are also in debian themselves 19:45
but last (to me) known state was that libtommath hasn't had a release for ages, which is why Debian didn't update 19:46
and maybe we wanted to avoid diverging too much from last release?
robertle ok, I think I kinda understand the current state now 19:48
and yeah, it is totally about the non-fat-packaging in debian :) 19:49
vrurg Are annotations on QAST purely informative? I'd like to pass on some info for the optimizer and they look like the thing to use. 20:02
AlexDaniel lizmat: that's correct, moving repos one by one will work better (even if it sounds a bit surprising) 20:20
lizmat: I'll work on it after the release
lizmat: meanwhile you can say if you like github.com/rakulang or github.com/raku-lang more 20:21
I guess rakulang?
lizmat yes, as it is the same as for social media
El_Che is it too early for Perl7 jokes? :)
lizmat OTOH, /raku is not an option / 20:22
El_Che: yes
El_Che :)
lizmat it is always too early for that
AlexDaniel lizmat: there's a semi-active account, you can drop them a message and ask what they think
[Coke] I have a slight preference for raku-lang
El_Che [Coke]: really? I am surprised 20:23
[Coke]: you ain't no UNIX programmer, mister. That's one char more for typing :) 20:24
==> rklng
20:27 patrickb joined
AlexDaniel fwiw it was registered as RakuLang, and it matters 20:29
[Coke] I'm not much of anything any more, but approaching 50, there's something to be said for a little bit of clarity
AlexDaniel it's case-insensitive but it'll default to how it was registered any time you click away 20:30
lizmat: but really, do drop them an email, there's a tiny chance that it'll save us some bikeshedding 20:31
El_Che hehe, a new one: NQP::Config::Rakudo=HASH(0xd6c718), "Cannot determine the brand of your make utility." 20:32
Already falling in love with Windows :)
and whoever wrote that error message, is very polite
lizmat so, how do you send a message to the owner of a repo ? 20:38
AlexDaniel: ^^ ? 20:39
AlexDaniel lizmat: pm
tbrowder AlexDaniel: I queried travis-ci about increase in build time and just got response from a Jan Schulte, Solutions Engineer, saying they would be happy to extend the time and asked for what repos. I assume rakudo/rakudo should be the one to start with to see how it goes? 20:42
AlexDaniel El_Che: ↑ 20:44
rba: ↑
patrickb El_Che: I at least found the error with 'bash' not found 20:48
tellable6 2019-11-08T17:34:07Z #raku-dev <El_Che> patrickb: gist.github.com/nxadm/772b514ea812...8f9671dbbf
patrickb Still no luck with reproducing the relative stuff in path problem.
tbrowder I did tell them it would take up to two hours but they didn't seem to blink. They can also increase to more powerful virtual machines but they would need to talk about options via phone call, "10 to 15 minutes." Number: +49 151 22353618 e-mail: <jan@travis-ci.com> 20:49
AlexDaniel ok maybe it's not the best idea to share that publicly? :) 20:50
[Coke] oof. yah, I wouldn't share my cell on public IRC, eek. 20:51
El_Che tbrowder: for the rakudo-pkg repo I split everything in parallel tasks 20:55
tbrowder: however, for i386 performance is tricky and the builds often fail 20:56
tbrowder: it's however not a repo in the rakudo namespace, so probably not relevant in the discussion 20:57
patrickb El_Che: I pushed the fix for the 'bash' not found problem. The problem is obvious in retrospect... 20:59
tbrowder sorry, i don't worry about sharing mine, so i didn't think about not sharing a business contact
El_Che patrickb: oh? 21:00
patrickb The script relies on $0 to contain the script call. When sourcing the script that doesn't work
So I switched over to `set-env.sh` and made set-env.sh print the `export PATH=...` stuff. 21:01
El_Che patrickb: I went pretty much the other way: the $0 way (github.com/nxadm/rakudo-pkg/blob/r...o-path#L7)
ah, interesting
patrickb Yeah, that works in restricted usage scenarios as yours. Determining the script path is awefully non-trivial when done correctly. The code in that set-env.sh script is the absolute shortest I found that doesn't break in most corner cases... 21:04
El_Che patrickb: can cmake used for building rakudo on Windows? It's preinstalled on travis. Or is nmake the way to go (looking at your instructions) 21:05
patrickb El_Che: CMake doesn't work. That's a completely different pair of shoes.
nmake is the way to go.
El_Che thx, some time saved
patrickb you should have nmake installed along side the vc compiler anyways...
El_Che is nmake part of visual studio?
patrickb yup 21:06
also in the Visual Studio Build Tools for that matter.
El_Che yeah, stuff is preinstalled but not easy to figure out where (running find on travis no no fun :) )
patrickb El_Che: I bet your "Relative paths in PATH with set-env.sh" problem stemed from calling it like ". ./set-env.sh". It probably picked up the first "." as $0 somehow and went from there. 21:08
El_Che patrickb: it makes sense, indeed
patrickb Then I'll be back to tuning the download pages...
El_Che patrickb: last question: any specific VS version specifically needed? 21:09
is 2017 ok?
patrickb El_Che: In all my tests I used latest (2019) but I think older ones should work fine too. 21:10
El_Che ok thx 21:11
patrickb (Fun Fact: The latest update of VS 2019 as of a week ago has a bug that makes the compiler fail during MoarVM compilation...)
El_Che oh 21:16
patrickb: ok, nmake added to te path, it start building. We'll see where it explodes :) 21:40
You people are all anti-Windows: 21:41
Cannot continue after this error.
On linux, maybe you need something like 'sudo apt-get install build-essential'.
On macOS, maybe you need to install XCode and accept the XCode EULA.
only linux and macos tips :) 21:42
21:48 travis-ci joined
travis-ci Rakudo build failed. Elizabeth Mattijsen 'Slightly other approach to handling "a".UInt 21:48
travis-ci.org/rakudo/rakudo/builds/609403050 github.com/rakudo/rakudo/compare/a...30cc76a5c8
21:48 travis-ci left 22:14 MasterDuke joined
MasterDuke moritz: yeah, libtommath just released v1.2.0 github.com/libtom/libtommath/blob/...anges.txt. archlinux has updated its package, but ubuntu (and i assume debian) hasn't yet 22:19
i think i tested building with a stock v1.2.0 (i.e., without trying to rebase the commits added to the fork we use in moarvm) and everything built+passed spectests 22:20
but even though timotimo has told me how a bunch of times, i still have no confidence in my knowledge of how to make a PR for the update 22:21
22:54 MasterDuke left
tbrowder .ask AlexDaniel what shall i say to Jan Schulte? 22:55
tellable6 tbrowder, I'll pass your message to AlexDaniel
AlexDaniel .
I don't know! 22:56
which is why I pinged El_Che and rba :)
El_Che: what do you think?
El_Che AlexDaniel: It depends what you want to do on travis. Faster VMs could speedup rakudo builds after pushs, but whatever you do the builds will take some times. Also, I think I am the only one using travis for releases. 23:01
AlexDaniel El_Che: do you struggle with timeouts? 23:02
El_Che AlexDaniel: it's better to think hard before asking stuff, because it's better to ask what you need the first time
AlexDaniel: timeouts? Only with Blin
not with rakudo itself
AlexDaniel hmm
El_Che I have around 15 parallel jobs for a release
maybe more
AlexDaniel shrugs and runs away
El_Che 22 actually :) 23:03
each job is a different distro
23:04 patrickb left
El_Che The problem I had with Blin is more "our fault" (the Digest module or zef): travis kills a job if it doesn't produce any output for a long time (like 10 minutes) 23:04
AlexDaniel nah, I'm convinced travis is not going to make it easier to run blin in the long run 23:06
the ecosystem is getting bigger so we'll gradually need bigger timeouts
El_Che AlexDaniel: no, not for running Blin, for creating an image 23:07
AlexDaniel hmm
El_Che For blin, we can get VMs
AlexDaniel right
ok ok
El_Che e.g., digitalocean sponsers a friends project
AlexDaniel runs away :)
El_Che (but if travis gives you a big machine without timeouts, you could run blin there :) ) 23:09
23:24 dogbert17 left