genpaku has quit [Remote host closed the connection]
genpaku has joined #crux-devel
SiFuh has quit [Remote host closed the connection]
SiFuh has joined #crux-devel
<beerman>
jaeger: https://dpaste.com/6R9ESBTKU it seems that docker depends on btrfs-progs (?), but even with it, it fails to build for me on one machine
<beerman>
oh, it already does pull in btrfs-progs, silly me. sorry. the paste is still relevant. I am currently trying inside my container
<jaeger>
yeah, discovered that last night. Recent btrfs-progs update broke it, maybe? I haven't looked into it yet
<jaeger>
just tested here, downgrading btrfs-progs to 6.0.2 fixes the docker build
<jaeger>
Based on the comments in that issue I don't think the docker folks are rushing to fix it, so it might be a while. We could either revert the btrfs-progs update or disable btrfs support in docker temporarily, perhaps.
<jaeger>
Looking at yet another ffmpeg/kodi breakage for kodi 19.5 (which I know you already work around by using an older ffmpeg tarball) and I'm wondering if we should introduce a Pkgfile option to skip unpacking specified things in the source array for situations like this.
<jaeger>
Instead of overriding unpack_source, which feels messy even though I've also done it a few times
<jaeger>
We've probably talked about this before, I don't recall
<farkuhar>
jaeger: the renames array can be used for that purpose, as seen in this diff http://ix.io/4jOx
<farkuhar>
This tactic might be criticized for violating the KISS principle, by "hiding one function inside another". It does avoid having to add new subroutines to pkgmk, though.
<jaeger>
I'll take a look, though from a logical standpoint I think I'd rather handle it in unpack_source, no new subroutine
<jaeger>
It would just be something simple like "if this source filename exists in nounpack=(), then just copy it instead of unpacking it"
<farkuhar>
The original unpack_source function already makes the distinction between files that get copied intact, and files that get unpacked. It just uses the crude mechanism of filename extensions to make that distinction. So I figured, let's play along with its crude method, and give it a filename that doesn't match what it likes to unpack. Then we can tidy up the mess on the other side of unpack_source (in the build function).
<jaeger>
Yeah, looking at SKIP I'm not a fan of it for this... it works but it's not intuitive and requires specifying the entire source array as renames, seems like
ivandi has quit [Quit: WeeChat 3.7.1]
ivandi has joined #crux-devel
<beerman>
including the entire source array is kind of a hassle with a few things, i agree, but its the best tool we have so i'm gladly using that one
<beerman>
about btrfs-progs and docker.. dunno, might sense to wait a bit. docker folks seem a bit odd, at times.
<jaeger>
I'm not saying we shouldn't use SKIP, I just don't like it for this particular use case. I'm going to play with another solution to see how I like it
<beerman>
sure, i'd take anything better in a heartbeat
<ivandi>
imho downgrading btrfs-progs is a better option than disabling btrfs. we don't lose functionality in docker. on btrfs filesystem docker uses the btrfs storage driver by default.
<jaeger>
playing with a "nounpack" array currently
<beerman>
where is that?
<beerman>
python?
<jaeger>
pkgmk/Pkgfile
<jaeger>
Ironically I did add a single subroutine for this to check if a value exists in an array, but it could be a one-liner instead of its own function if that really annoys people
<beerman>
true 🙂 but it failed once for a day, might make you wonder when elon aquires github and takes that particular server down that does it, or whatever else might happen 😄
<jaeger>
I only use it to get patches, not in Pkgfiles
<beerman>
you don't use it to rename distfile names?
<jaeger>
Not the .patch functionality, which is what I assumed you were talking about since your comment came right after that
<beerman>
ah, well, patches i'd download and proably rename too
<ivandi>
beerman: weechat build fails with ruby-3.2. sed -i 's/ruby-3.1/ruby-3.2 ruby-3.1/' $name-$version/cmake/FindRuby.cmake fixes it.
<beerman>
thanks, will look into it tomorrow, probably
<ivandi>
np
<jaeger>
yeah, that's what I'm doing with a couple of patches now... one for kodi and one for docker