-
-
Notifications
You must be signed in to change notification settings - Fork 14k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
default-gcc-version: 12 -> 13 if !isDarwin && !isAarch64 #268097
Conversation
Note: the builds ( |
I've built the entire @vcunat I would like to merge this to staging once 23.11 has branched off. |
I've started building the entire, full release jobset ( Not sure how long this will take. I'm using my own tooling instead of the Hydra software, and it tends to get stuck a lot. We'll see how it goes. |
riscv64 is failing due to missing atomic ops in gcc 12. This commit works around the problem by linking against libatomic. It can be reverted when we switch to gcc 13 (NixOS#268097)
After many
When was the last time an entire release jobset was built by anything other than |
riscv64 is failing due to missing atomic ops in gcc 12. This commit works around the problem by linking against libatomic. It can be reverted when we switch to gcc 13 (#268097)
Alright, I'm running one last pass of the leftovers (~1,723 jobs) and I'll triage what's left after that. |
env = lib.optionalAttrs (stdenv.cc.isGNU && lib.versionAtLeast stdenv.cc.cc.version "13") { | ||
NIX_CFLAGS_COMPILE = "-w"; | ||
}; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do they pass Werr? If so we should remove that instead
I have filtered the following derivations which are FODs for no-longer-available (and not on
|
Try reverting this on a future ROCm release. Addresses: #271943
|
This pull request has been mentioned on NixOS Discourse. There might be relevant details there: https://discourse.nixos.org/t/prs-ready-for-review/3032/3058 |
Does it build now? |
Indeed it did!
Indeed it does!
|
This broke
Traced back to
which sets
|
Possibly more packages to look into
|
This pull request has been mentioned on NixOS Discourse. There might be relevant details there: |
https://nixpk.gs/pr-tracker.html?pr=268097
Did this PR break your package? Please read before posting.
Please accept my apologies.
This PR is expected to break up to 321 out of the 16,938 packages in nixpkgs. I fixed all of the non-leaf packages, but there were simply too many for me to handle all of the leaf packages by myself.
Before posting below, please try these easy fixes:
disable-warnings-if-gcc13
around your package. Example:https://github.com/NixOS/nixpkgs/blob/9855b101b3f2e7759ca8daaa9ff6cca1c90a4a7d/pkgs/top-level/all-packages.nix#L12535-L12537
pin-to-gcc12-if-gcc13
around your package. Example:https://github.com/NixOS/nixpkgs/blob/9855b101b3f2e7759ca8daaa9ff6cca1c90a4a7d/pkgs/top-level/all-packages.nix#L2792
Thanks, and again, sorry about the breakage.
0900 UTC
today (2023-Dec-07
) unless somebody objects. I don't think anybody else has the compute to test this PR, so I will take all the blame for any breakage. See next section for what I expect.Testing against
staging
(i.e. "as-will-be-merged")Note: a "job" means one top-level attrpath, like
python3Packages.pycairo
, on one platform (currentlyx86_64-linux
). This corresponds to one derivation and many outpaths. Not every derivation corresponds to a job; there are about 115,167 derivations in the closure of the 16,938 jobs' derivations -- I think Hydra counts every derivation in the closure as a "job".For the
release-combined
jobset forx86_64-linux
built at 195e990 as ofThu Dec 7 05:56:55 UTC 2023
Not all 321 failures are due to this PR -- that number includes any packages that already failed to build on
x86_64-linux
as of 7629c42. The rest will require individual attention from their maintainers. I applied fixes (either-Wno-error
or pinning togcc12Stdenv
) to all of the non-leaf packages that were broken. What remains is going to require ~321 individual fixes from the maintainers of those packages, and I'm getting a little bit burned out.In a few cases (mainly
deepin
) the package expressions were so complicated that I could not figure out how to properly override withstdenv = gcc12Stdenv
. Big thanks to @Madouura for doing most of that (and showing me how to do the rest) forrocmPackages
.Job lists as of
Thu Dec 7 03:12:05 UTC 2023
:Earlier testing against
release-23.11
branchMore than 99% of the full release jobset builds correctly with this PR (less than 1% fail), so I think it's getting to the point of diminishing returns for pre-merge fixes... right now
staging
/master
are starting to diverge faster than I can fix things. So I would like to proceed with merging this.release-small
jobset has built successfully on Shesha.release-combined
jobset is completeOut of 51,255 derivations, using
release-23.11
as the base branch:List of failed derivations
Will need Hydra for the Darwin builds.
Includes:
Description of changes
Things done
x86_64-linux
tests.cross.sanity
x86_64-linux
workstation+serverpowerpc64le-linux
workstation+serveraarch64-linux
workstation (cross fromx86_64-linux
)mips64el-linux
router (cross fromx86_64-linux
)