00:12 kjp left 00:14 kjp joined 00:15 kjp left, kjp joined 07:20 Guest50 joined 07:21 Guest50 left 08:24 sp1983 joined 08:49 sp1983 left 12:21 Pixi left 12:36 Pixi joined 15:19 sp1983 joined
sp1983 o/ 15:20
there is error when building Hash::Merge against Rakudo 2026.03 in aarch64 - lang-call cannot invoke object of type 'VMNull' belonging to no language
at /builddir/build/BUILD/raku-hash-merge-2.0.0/vendor#sources/3A740CC2EDEBFDBE342C298E66C87C9E3533CDFA (Hash::Merge::Augment):6
does it ring any bell to any one ? 15:21
here is full log - download.copr.fedorainfracloud.org...ive.log.gz
^^ El_Che:
lizmat builds ok on MacOS 15:22
fwiw
15:26 sp1983 left 15:28 sp1983 joined
sp1983 yeah, same for me 15:28
disbot2 <melezhik.> Probably need to build against fresh aarch64 binary myself 15:36
lizmat is it repeatable, or a flapper ? 15:39
sp1983 sorry, for confusion it happens on x86_64 15:44
looks like a flapper 15:45
based on "I've tried building in a couple of different ways, and under some conditions it does not complain.." 15:46
15:56 sp1983 left 16:00 sp1983 joined
sp1983 or better say environment dependent 16:01
16:02 grayeul joined 16:08 sp1983 left, sp1983 joined 16:12 sp1983 left 19:09 [Coke] left 21:05 [Coke] joined
grayeul So ... I have a sort of packaging question... I'm not very familiar with raku usage, but I've been asked to put together some rocky-linux rpm packages to provide a controlled set of rakudo baseline packages. 22:18
lizmat El_Che patrickb timo ^^ 22:20
grayeul I'm trying to install (using install-dist.raku) a few different packages (individually -- each into their own rpm). What I'm seeing is that two different packages are 'fighting' over the same precomp files. So, I'm trying to understand why when I install (e.g.) Terminal::ANSIColor I get some of the same files as when I install JSON::Fast
For example: hare/perl6/vendor/.precomp/CD2858E4B48D26E12AB7903CDD61A7201262E08D/4C/4C8907699141DCFF1B0B1A3B6A4D49D076402168.repo-id 22:21
I don't know how to tell what generates that particular file, and why it would appear in two different packages. -- any help appreciated...
I also have a different problem (mentioned earlier) .. when trying to build the Hash::Merge into an rpm. I think the problem relates to the fact that I'm installing into a non-standard location (within a /buildroot) that won't be where the module appears in the end... but it seems during the 'prep' of the Hash::Merge package, in the Hash::Merge:Augment area, there is a 'use Hash::Merge' -- and I think 22:24
it is not finding that, as I'm still trying to install it (and it isn't in an expected place).
I tried setting RAKUDOLIB to add the path I'm putting things, but that didn't seem to help... I *think* that just adds to the standard path, right? I don't have to list all possible paths if I do set RAKUDOLIB 22:25
timo salsa.debian.org/perl6-team/dh-rak...type=heads steal from how debian builds the modules, or how they are built for opensuse: build.opensuse.org/projects/devel:...c?expand=1 22:55
the RAKUDO_RERESOLVE_DEPENDENCIES=0 might be relevant
grayeul I did have that set to 0 23:19
timo OK, with RAKUDO_MODULE_DEBUG you should see mentions of the files that conflict and that should help see where it comes from 23:22
grayeul I'll try that... but there is no conflict when I build pkg A, or pkg B separetely -- it is at install time that 'rpm' complains two different packages are trying to provide the same file. 23:26
Is the .precomp directory needed? Looks like that's where the conflicts are... 23:30
timo I've encountered issues when doing debian packaging where a not-tight-enough dependency version specification caused the package with the dependency to separately precompile the dependent package even though it should have taken the existing one because the version from the specification didn't match the version of the dependent package in a check somewhere 23:35
maybe that's what's happening to you, but I don't think Terminal::ANSIColor has any dependencies, and I know that JSON::Fast doesn't
so the only thing it might be depending on and causing a re-precompile of could be parts of rakudo itself 23:36
I was able to figure that out from the RAKUDO_MODULE_DEBUG output
grayeul yeah, there shouldn't be any real dependencies in play, I don't think... that's why it is odd -- 23:57