Summary: | net-misc/rsync fails to cross compile for musl (CBUILD=x86_64-pc-linux-gnu, CHOST=x86_64-pc-linux-musl) | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Kirill Chibisov <contact> |
Component: | Current packages | Assignee: | Gentoo's Team for Core System packages <base-system> |
Status: | UNCONFIRMED --- | ||
Severity: | normal | CC: | grobian, musl |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | AMD64 | ||
OS: | Linux | ||
See Also: | https://bugs.gentoo.org/show_bug.cgi?id=732084 | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
Compilation failure log
Configure output |
Description
Kirill Chibisov
2022-03-12 01:44:18 UTC
Created attachment 766806 [details]
Compilation failure log
FWIW, the fact that this is cross-compilation is quite relevant I think, given that we've had a loooot of bugs related to that w/ rsync + SIMD, even outside of musl. what confuses me: khnum% emerge --info | head -n1 Portage 3.0.30 (python 3.9.9-final-0, default/linux/ppc64/17.0/musl, gcc-11.2.0, musl-1.2.2-r7, 5.15.23-gentoo ppc64) khnum% qlist -IvR rsync net-misc/rsync-3.2.3-r5::gentoo sokaris% emerge --info | head -n1 Portage 3.0.30 (python 3.9.9-final-0, default/linux/arm64/17.0/musl, gcc-11.2.0, musl-1.2.2-r7, 5.15.23-gentoo aarch64) sokaris% qlist -IvR rsync net-misc/rsync-3.2.3-r5::gentoo None of these are cross, but can you show your configure output? For it looks like it wrongly determines it can use the simd instructions while it shouldn't. I don't have x64 platform using musl, so that may be a reason too. Created attachment 766848 [details]
Configure output
My host has all simd stuff, my target won't have it though. The only simd stuff it'll have is sse2. I was able to build rsync from ::gentoo without any issue in chroot. So it looks like cross issue indeed. hmmm, the cross patch is supposed to address, exactly this. I wonder what happen without that patch applied. It replaces build with host, but I wonder if it shouldn't use target instead. (In reply to Fabian Groffen from comment #7) > It replaces build with host, but I wonder if it shouldn't use target instead. Unlikely; we are not building a cross compiler here. |