🦋 Welcome to the IRC channel of the core developers of the Raku Programming Language (raku.org #rakulang). This channel is logged for the purpose of history keeping about its development | evalbot usage: 'm: say 3;' or /msg camelia m: ... | Logs available at irclogs.raku.org/raku-dev/live.html | For MoarVM see #moarvm Set by lizmat on 8 June 2022. |
|||
00:25
donaldh left
03:12
guifa left
04:40
guifa joined
|
|||
nine | ab5tract: oh my....of course that my $_ in RakuAST::Call::Method.IMPL-INTERPRET would not help. We're looking for a $_ in the lexical outer frames, not the caller frames. | 07:57 | |
Ah, no. That's not it. As EVAL does my $context := nqp::defined($ctx) ?? $ctx !! CALLER::LEXICAL::; | 07:58 | ||
08:52
finanalyst joined
09:34
finanalyst left
|
|||
ab5tract | What I’ve noticed is that the blow up is happening with a compile resolver, presumably being handed off from the eval resolver | 10:24 | |
But I was a bit surprised to be working with a compile resolver where I would have expected an eval resolver | 10:25 | ||
nine: any idea what RakuAST::IMPL::InterpContext is for? It’s passed to all IMPL-INTERPRET calls but is an empty class and I never see anything done with it | 10:26 | ||
I *think* the issue could be solved by correctly passing the eval resolver’s outer to the compile resolver constructor. But I’m really just stabbing at shadows at the moment | 10:27 | ||
nine | ab5tract: resolvers wont help you | 10:48 | |
The EVAL resolver is only for evaluating full RakuAST trees which is a different use case | 10:49 | ||
That context is also unused for now | |||
ab5tract | that saves me from wasting time going down false paths. thanks! | 10:51 | |
I do wonder about that though.... `src/Raku/ast/resolver.rakumod:1004` appears near the top of the stack trace | 10:52 | ||
But also: taking your word for it usually works out well for me :) | |||
If you have any thoughts as to where I can direct my future stabbings, please share! | 10:53 | ||
nine | It's something about CALLER::LEXICAL not finding symbols from the nqp code. IOW I don'tknow how to give it a $_ | 11:00 | |
You can work around in EVAL itself | |||
Check whether $ctx has a $_ and if not take a LEXICAL:: froma block as $ctx | 11:01 | ||
I would really like to know if theres another way though. PseudoStash is not easy to understand though | 11:02 | ||
ab5tract | I think the from-context code is wrong because it disregards the resolver’s $!outer when constructing the new derived resolver | 11:23 | |
nothing changed when adjusting this, though :( | 11:24 | ||
I'll try the less awesome approach first | |||
Wow, you were not joking about PseudoStash :O | 11:31 | ||
11:46
El_Che joined
|
|||
El_Che | timo: test mimalloc_alpine_fix_for_2025.04_release atm | 12:00 | |
12:01
guifa left
|
|||
El_Che | bad news: | 12:03 | |
productionresultssa18.blob.core.wi...4-997b-47e | |||
9-b12b-9515b896b4de&skv=2025-01-05&sp=r&spr=https&sr=b&st=2025-04-21T12%3A03%3A07Z&sv=2025-01-05 | |||
ooops | |||
to long url | |||
summary on top, full log below: gist.github.com/nxadm/4109ec01e6cc...a8addcf0d3 | 12:04 | ||
timo | is there a package that has the linux headers inside it? | 12:20 | |
i've seen a different patch where the include for prctl was moved inside the if(__GLIBC__) below, could you try that? that'd be the "if !defined(PR_SET_VMA)" and also the include for prctl.h | 12:51 | ||
github.com/microsoft/mimalloc/comm...3518f5dde4 - so this patch that's already included from the moarvm branch, but moved one line down essentially | 12:52 | ||
On musl-libc the required types for prctl are defined in sys/prctl.h and including linux/prctl.h fails either because the file does not exist or because it redefines the types. | 12:54 | ||
^- from one of the issues on github | |||
El_Che | Hi | 12:57 | |
yes, linux-headers | 12:58 | ||
and compiling with gcc | |||
I addedd it before | |||
before the branch, whilke debugging | |||
but it then complains a function is redeclared | |||
[Coke] | Maybe it's better to go back than forward here. | 13:08 | |
timo | the code below that also uses prctl uses "if defined" to make sure it doesn't try to call prctl with something that's missing | ||
yeah coke i think you're right | |||
[Coke] | And only update mimalloc until after it's safe. | ||
timo | El_Che: can you easily change the mimalloc 3rdparty commit for the submodule? or just revert the individual commit that updates the mimalloc version? | 13:09 | |
13:33
finanalyst joined
13:34
finanalyst left
|
|||
[Coke] | patrickb: do we have any docs describing our current CI? | 14:36 | |
seeing a bunch of warnings in dev.azure.com/Rakudo/rakudo/_build...ew=results | |||
patrickb | [Coke]: Sadly I don't. It's only in my and timo's head... | 15:01 | |
I believe many of those warnings relate to the release build pipeline. | 15:03 | ||
jdv | it is slightly odd we only really guarantee any rando release will work on vanilla linux distros | ||
patrickb | Those have the special requirement to run on the oldest possible OS. | 15:04 | |
jdv | might need to trade velocity for platform support i guess | ||
patrickb | That's because of glibc being forward, but not backward compatible. | ||
timo | ok so the warning about tools in the ubuntu image not working any more doesn't concern us i don't think | ||
patrickb | Thus I build on one of the distros with the oldest glibc still in use. | 15:05 | |
IIRC it's CentOS 6 in a docker container on an Ubuntu host. | 15:07 | ||
[Coke] | patrickb, timo: as time permits, getting that documented (wiki page in rakudo git repo?) would be great. Would have loved to have an early warning about the el_che issue for this release, e.g. | ||
jdv | id guess nobody else builds on alpine | 15:08 | |
timo | we could maybe get people who have their own packaging to contribute a little bit to the rakudo release pipeline on azure pipelines? most should be able to get their build stuff into a dockerfile and we could build that in our pipeline | 15:09 | |
well, linux distros that is | |||
jdv | i only build on debian:| | ||
[Coke] | my azure box is ubuntu | 15:29 | |
(for blin & release stuff) | |||
jdv | ?ironic? that m$'s cloud stuff is basically linux | 15:30 | |
[Coke] | oh, it's an option on the VMs - I could be using any of a number of windows boxes also | 15:32 | |
(looks like they also have some RHEL options, and then some one offs) | 15:33 | ||
jdv | even more odd is a friend of mine workscat amazon scripting azure stuff:) | 15:34 | |
timo | where all my comet snowman lovers at? | 17:20 | |
22:15
guifa joined
22:38
guifa left
|