-
-
Notifications
You must be signed in to change notification settings - Fork 14.4k
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
python3Packages: don't depend on setup hooks' bash eval (part with fewer rebuilds) #354811
Conversation
|
|
@ShamrockLee Does conan fail on master already or is this a new failure? |
Here is the build log of the failed build: |
Hm, can't reproduce it on aarch64-darwin, x86_64-darwin or x86_64-linux.
Could you try to rebuild this with Alternatively, could you try to apply just the change to conan's file on master - to make sure we're looking at the right change here and it's not a different change in another file that causes this? If the removal of |
I set I'll also build it on NixOnDroid, but that would take a while because my phone lacks CPU power. @ofborg build conan |
That was on |
The failure happens when building |
Oh! Now I get what you mean. I'll build it on |
The build turns out to be reproducible on the The |
The build log of the |
Wow, really odd. I am out of ideas right now 😕 |
@HaoZeke, what do you think about Conan's test failure? |
Ah, I misread your message earlier. So indeed the test is flaky for aarch64-linux on master already, right? In this case, I would suggest we just disable the specific test and then continue. |
The nixbuild.net aarch64-linux builder fails |
To be able to remove the bash-eval behavior for setupPyGlobalFlags, we change $out to use placeholder "out" instead. Confirmed that the locales are still loaded from the correct path as demonstrated in NixOS#284050.
There is only one test and this was disabled via some nasty bash eval hacks in pytestFlagsArray. Let's just use... doCheck = false instead?
We'd like to avoid bash eval in pytestFlagsArray, because we want to get rid of support for it. This case works just fine without.
It makes no sense to depend those folders on pname - they are part of the source code, not our convention, after all. Also this avoids using bash eval behavior in pytestFlagsArray.
This surely was intended to be an imports check, judging by the content of the list. Setting pytestCheckHook like that surely makes no sense.
Those are provided by pytest-xdist's setup hook automatically anyway.
This flag relies on bash eval of pytestFlagsArray, which we'd like to get rid of. By moving the evaluation of $src into the preCheck hook, we achieve the same.
We can just cd to $out to let the tests run from that folder automatically. Additionally we get better test output, because the /nix/store component is missing from file names.
Avoids the use of bash eval in unittestFlagsArray, which we want to remove.
This test failed repeatedly on aarch64-linux, so far.
29cc4a5
to
f9969cb
Compare
I disabled the @ShamrockLee Can you test with nixbuild.net again? |
@ofborg build conan |
Now |
Conan built on aarch64-linux for OfBorg successfully as well. This should be good to go, imho. |
This is a part of #352976. Those packages here should cause almost no rebuilds, so targeting master.
@ShamrockLee @emilazy
Things done
nix.conf
? (See Nix manual)sandbox = relaxed
sandbox = true
nix-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.