linux: fix compiler selection#402198
Conversation
RossComputerGuy
force-pushed
the
fix/kernel-llvm
branch
from
32ec98b to
8818b45
Compare
github-actions
bot
added
10.rebuild-darwin: 1-10
nix-owners
bot
requested review from
K900,
Ma27,
NeQuissimus,
TredwellGit,
alyssais and
thoughtpolice
|
I'd suggest keeping Would be good to have the two "verify" commits from the other PR as well. |
|
Tried rebasing the original with the My hack of removing |
RossComputerGuy
force-pushed
the
fix/kernel-llvm
branch
from
8818b45 to
35477af
Compare
|
Cool, this seems to be better. |
RossComputerGuy
force-pushed
the
fix/kernel-llvm
branch
from
35477af to
9f2e5bb
Compare
I got progress, not sure about this error. Apparently, I had forgotten a couple things which made things not actually build with LLVM but it is now and it's failing here. |
There was a problem hiding this comment.
This needs to be lifted somewhere, and reused. We can't have this in two places at once and risk getting out of sync.
There was a problem hiding this comment.
Where should that be?
There was a problem hiding this comment.
I don't know, but it should be in one place and not two.
There was a problem hiding this comment.
This is done now.
There was a problem hiding this comment.
Yeah, I like this a lot more now.
RossComputerGuy
force-pushed
the
fix/kernel-llvm
branch
3 times, most recently
from
342687b to
ac4b372
Compare
|
|
|
Built pkgsLLVM.linux on aarch64 and x86_64. |
|
Anything blocking this to merge or can we just merge it? |
RossComputerGuy
force-pushed
the
fix/kernel-llvm
branch
2 times, most recently
from
46845bd to
ddf6260
Compare
This reverts commit 70cc251. This commit caused gcc to be pulled in as a target compiler for pkgsLLVM. Co-authored-by: Tristan Ross <[email protected]>
Co-authored-by: Tristan Ross <[email protected]>
RossComputerGuy
force-pushed
the
fix/kernel-llvm
branch
from
ddf6260 to
c242028
Compare
|
Whoops, broke cross. Fix should be #429588 — still building to confirm. |
|
At least |
|
Right — userspace tools will need to continue using the wrapped compilers. |
|
perf fix: #429818 |
|
FYI: I want to make you aware that this PR causes a regression for at least zfs and possibly v4l_loopback. I was trying to bisect the build error for zfs. The bisect on zfs_2_3 build this result e9b2edb in being the first bad commit, double checked that a155990 (parent) still builds Havent checked for v4l_loopback in more detail (just found it randomly by clicking hydra logs). But the logs look quite similar to me |
Since NixOS#402198 kernel modules are no longer built with wrapped compilers. Firstly, NIX_CFLAGS_COMPILE is no longer effective, so we change that to CFLAGS, which is used in the Makefile for evdi. Secondly, we still need to use the wrapped compiler for userspace stuff, so separate that out in postBuild. Fixes NixOS#437311
|
|
Since #402198 kernel modules are no longer built with wrapped compilers. Firstly, NIX_CFLAGS_COMPILE is no longer effective, so we change that to CFLAGS, which is used in the Makefile for evdi. Secondly, we still need to use the wrapped compiler for userspace stuff, so separate that out in postBuild. Fixes #437311
9 participants
Things done
Based on #390631. Thank you @blitz for doing a good chunk of the work. And thanks to @zowoq for finding the fix for Rust. Best way to verify this is by building
pkgsLLVM.linux.configfileand checking ifCONFIG_CC_IS_CLANGis set, around that same area is should reference clang v19. Try buildinglinux.configfile(non-pkgsLLVM) derivation and see if it references gcc still. Ofc, it's good to verify the full kernel build works. I tested both of these on my system and it works as expected.nix.conf? (See Nix manual)sandbox = relaxedsandbox = truenix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage./result/bin/)Add a 👍 reaction to pull requests you find important.