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
bazel: 1.2.1 -> 2.0.0 #76851
bazel: 1.2.1 -> 2.0.0 #76851
Conversation
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.
diff LGTM (did not check thoroughly)
bazel --help
works
nixpkgs-review
doesn't pass
Missing dependencies:
python3.7-annoy-1.16.3
:h5py
python3.7-gym-0.15.4
,python3.8-gym-0.15.4
:opencv-python
Check failures:
python3.7-minio-5.0.5
:
File "/build/minio-5.0.5/minio/helpers.py", line 292, in is_valid_endpoint
raise InvalidEndpointError('Hostname cannot have a scheme.')
minio.error.InvalidEndpointError: InvalidEndpointError: message: Hostname cannot have a scheme.
----------------------------------------------------------------------
Ran 133 tests in 0.037s
FAILED (errors=80)
Test failed: <unittest.runner.TextTestResult run=133 errors=80 failures=0>
error: Test failed: <unittest.runner.TextTestResult run=133 errors=80 failures=0>
@GrahamcOfBorg build bazel.tests |
@GrahamcOfBorg build bazel.tests.protobuf |
Looks like all tests pass. Merging. |
This broke the I would also expect to keep both versions of the package. We are talking about the build system here, not some random leaf package where major version update is most probably fine. |
We can’t block on tensorflow when merging bazel updates, because that’s a much too complex package (that is to say from what I hear it’s a shitfest to maintain). We couldn’t add it to downstream tests in the first place anyway, because of the nonfree blobs.
Bazel just changed their versioning, they are going to release a new major version every three months for the foreseeable future as far as I can see. nixpkgs is a snapshot, so if you need a workling tensorflow you can pin to an older version of nixpkgs until the tensorflow maintainers fix it to work with the current bazel. If you really need tensorflow to work on master, you can revert this patchset on top of e.g. a master checkout or a fork. |
I think we can, and we should. It is a bit frustrating that tensorflow breaks every couple of months (hopefully now with the new release cadence every couple of months) because yet another bazel update was merged without testing tensorflow. That said, this is a bit of an outlier since this time its actually |
Well, I (personally) certainly cannot. But if you want, I can add you as maintainer to bazel so that you are pinged and can test it before we merge. |
Unfortunately I'm not willing to take on that responsibility alone either. I already probably shouldn't be in tensorflow's maintainer list, as I only maintain it on an as-needed (as in when I personally need it) basis. Still, I think we, as a community, should. Tensorflow is the biggest consumer of bazel in nixpkgs. In fact, its close to the only one. Its also very popular. Hopefully someone interested in keeping tensorflow running will take you up on the offer and add themselves to the bazel maintainer list. They can then test tensorflow on new bazel releases, but they should get reasonable time to do so before merging a new bazel release. |
I’m afraid I must echo @justinwoo’s sentiment here: https://twitter.com/jusrin00/status/1221371774497181698 The “community” is just people doing things because they want to, everything else is either paid or abuse. I’m maintaining bazel in nixpkgs for https://github.com/tweag, I’m sure we can help with tensorflow, too, for reasonable rates. |
Can't you guys just have a "stable" version of bazel for tensorflow, and the updated bazel? |
I agree with @deliciouslytyped. Just because Bazel releases a new version that doesn't mean everyone should stop what they are doing and update to the latest Bazel version. Is there a reason we can't have multiple bazel versions? I mean we've done that for others as well and it would probably not hurt allowing consumers of nixpkgs use older versions of Bazel. Especially if Bazel (rightfully) breaks things on the next major version it might be wise to just keep the old version around. For how long? I don't know. Does it hurt us carrying multiple versions? Probably not, we just provide the same loose guarantees as we currently do. Stuff might just not be use able (as it is right now). |
If we know exactly which version tensorflow works with, I guess we could keep one around for that. It would have to be maintained by tensorflow maintainers though. |
Now we just need someone that is willing to do that work.. I am certainly not. Already have my fair share of pain every day. :) |
This pull request has been mentioned on NixOS Discourse. There might be relevant details there: |
I'm willing to do the work, but I'm having trouble building bazel when I cherry pick 4fdea73 into master.
Here's my very sloppy prototype branch for adding a bazel 1 https://github.com/mjlbach/nixpkgs/tree/bazel_1 |
I briefly looked into this and also ran in the above error. I've also tried downgrading GCC but that also didn't really help :/ |
_TF_MIN_BAZEL_VERSION = '0.24.1' that's for 1.15.2 |
Yes, but it does build correctly on 4fdea73 with 1.2.1 |
yeah |
Motivation for this change
Things done
sandbox
innix.conf
on non-NixOS linux)nix-shell -p nixpkgs-review --run "nixpkgs-review wip"
./result/bin/
)nix path-info -S
before and after)Notify maintainers
cc @