2025-04-01 00:59:35 <DWwanghao> You're welcome :)
2025-04-17 21:42:02 <ncopa> i got a build error on loongarch64 when building wasi-libc:
2025-04-17 21:42:05 <ncopa>     7 |         if (n > BULK_MEMORY_THRESHOLD)
2025-04-17 21:42:06 <ncopa> 1 error generated.
2025-04-17 21:42:06 <ncopa>       |                 ^
2025-04-17 21:42:06 <ncopa> libc-top-half/musl/src/string/memset.c:7:10: error: use of undeclared identifier 'BULK_MEMORY_THRESHOLD'
2025-04-17 21:42:06 <ncopa> make: *** [Makefile:732: build/wasm32-wasi/libc-top-half/musl/src/string/memmove.o] Error 1
2025-04-17 21:50:24 <ncopa> hum
2025-04-17 21:50:30 <ncopa> it happens on x86_64 as well
2025-04-18 01:05:58 <DWTest1231> hi ncopa,I'm currently on vacation until April 24. I'll prioritize fixing the wasi-libc build failure issue as soon as I return from my leave.
2025-04-18 06:55:48 <cely> I wonder if it's too late in the 3.22 bootstrap process to re-enable dt_relr for Loongarch
2025-04-18 06:59:55 <ncopa> cely: is there any open MR?
2025-04-18 07:00:00 <cely> Anyway, i found out why wasi-libc is in the dep chain of abuild, because git has glib in makedepends, and glib has util-linux, and util-linux has asciidoctor, which is written in ruby and requires cargo to build
2025-04-18 07:00:01 <ncopa> or issue?
2025-04-18 07:00:16 <ncopa> oh, wow
2025-04-18 07:00:23 <cely> You enabled it before
2025-04-18 07:00:53 <cely> 4ebab75eb2fa
2025-04-18 07:01:19 <ncopa> and we disabled it again?
2025-04-18 07:01:30 <cely> That was with binutils 2.43
2025-04-18 07:02:50 <cely> Anyway, i remember you saying that you set BOOTSTRAP while bring up the stable builders, otherwise there'd be too many deps to build
2025-04-18 07:03:19 <cely> BOOTSTRAP would remove asciidoctor from util-linux makedepends
2025-04-18 07:03:46 <ncopa> ap recursdeps $pkgs | xargs ap builddirs | while read aport; do
2025-04-18 07:03:46 <ncopa>         APORTS_BOOTSTRAP=1 abuild -rk -C "$aport" || break
2025-04-18 07:03:46 <ncopa> done
2025-04-18 07:04:20 <ncopa> oh...
2025-04-18 07:04:21 <cely> Probably should be passed to ap recursdeps
2025-04-18 07:04:31 <cely> So it factors into dep calculation
2025-04-18 07:04:31 <ncopa> yeah
2025-04-18 07:07:49 <ncopa> it is just me who is stupid :)
2025-04-18 07:07:56 <ncopa> thanks cely for clearing that up
2025-04-18 07:08:20 <cely> You're welcome
2025-04-18 07:08:24 <cely> So, i guess we won't be getting apkv3 in Alpine 3.22?
2025-04-18 07:08:25 <ncopa> re dt_relr, I suppose we could enable it, but then I'd need to strat over with loongarch64
2025-04-18 07:09:05 <ncopa> this needs to pass https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/82593
2025-04-18 07:09:23 <ncopa> not sure if I need to rebuild toolchain and abuild for apk3?
2025-04-18 07:11:18 <cely> For dt_relr, no worries, if it can be enabled in edge (without having to wait for a stable release), we can probably do it later on before the next stable
2025-04-18 07:11:31 <ncopa>  Enabling dt_relr will cause Rust and related aports to fail
2025-04-18 07:11:31 <ncopa>     These flags are not fully supported yet
2025-04-18 07:11:36 <ncopa> so it was reverted
2025-04-18 07:11:41 <ncopa> is taht still the case?
2025-04-18 07:11:56 <cely> I don't think so
2025-04-18 07:12:10 <cely> I think 2.43.1 already cleared that up
2025-04-18 07:12:15 <ncopa> alright
2025-04-18 07:12:18 <cely> binutils 2.43.1
2025-04-18 07:12:19 <ncopa> lets re-enable it
2025-04-18 07:13:54 <cely> For apkv3, i think the plan was actually to have the apk index in both v2 and v3 formats (but not sure what happened in the recent meeting)
2025-04-18 07:14:02 <cely> For that, there would have to be changes to abuild
2025-04-18 07:14:23 <ncopa> the plan was to simply add apk3, which should be a "drop-in" replacement
2025-04-18 07:14:44 <ncopa> aafter 3.22 release we work on adding dual index
2025-04-18 07:14:50 <cely> Ok
2025-04-18 07:16:04 <ncopa> looks like it is only apk-tools-rs that fails
2025-04-18 07:16:12 <ncopa> i suppose we could merge it as is
2025-04-18 07:17:15 <ncopa> and then work on apk-tools-rs afterwards. seems like there are some open MR for it
2025-04-18 07:17:39 <cely> There are some issues with scripts that check $0 for script name
2025-04-18 07:17:47 <cely> or rather, package name
2025-04-18 07:17:49 <ncopa> ok?
2025-04-18 07:18:18 <cely> Like docbook-xsl
2025-04-18 07:19:03 <cely> but not sure how people would like to solve that, easiest would be to stop sharing the post-install script
2025-04-18 07:19:52 <cely> I brought this up with fabled a while ago, and there was some talk about passing pkgname in some other way, but not sure if that's been done
2025-04-18 07:20:33 <cely> From what i remember, the issue is the post-install script being executed from memfd now (if /dev/fd is available), so $0 no longer contains pkgname
2025-04-18 07:21:11 <cely> or rather, it does, but not in such a straight-forward way now (you can readlink it, i think)
2025-04-18 07:26:09 <ncopa> ok
2025-04-18 07:30:52 <ncopa> maybe we should discuss apk-tools in #alpine-devel
2025-04-18 07:31:45 <cely> There is also #apk-tools
2025-04-18 07:33:36 <cely> Anyway, i think not many aports are affected, probably only postgresql-timescaledb (shares post-install and post-upgrade) and docbook-xsl (shares post-install between 2 subpackages)
2025-04-18 07:35:39 <ncopa> yeah, I suppose its manageable
2025-04-18 10:48:37 <ncopa> cely: do you think we should include apk-tools3 in alpine 3.22? i'm almost done setting up the builders
2025-04-18 13:34:47 <cely> ncopa: i'm not sure. if we are going to use some of the new features for something, maybe; but here it seems it's only going to be a drop in replacement. i guess the question then becomes, how much of a drop in it really is..
2025-04-18 13:49:33 <cely> I've been trying out the v3 package format, and there's something i've realized, if i mess up some libraries that apk links to, making apk itself unusable, there is no more reaching for tar to fix it. the static apk binary will become much more important..for apk extract
2025-04-18 13:50:45 <cely> but all that's in the future. for now, i note that we haven't upgraded to apk-tools 2.14.10
2025-04-18 13:53:50 <cely> Hmm, apparently even that has some ABI change?
2025-04-18 13:54:00 <cely> !80006
2025-04-18 14:18:54 <ncopa> i think we will wait with the bump til after 3.22 release
2025-04-18 14:30:05 <cely> Ok
2025-04-18 14:44:21 <ikke> cely: fabled indicated there is no stable abi yet
2025-04-18 14:46:43 <cely> I think i was talking about 2.14.10, and the comment on that MR
2025-04-18 14:47:03 <ikke> Oh, sorry 
2025-04-18 14:47:40 <cely> (just wondering why apk-tools 2.14.10 has not been merged)