Revert "bazel: fix darwin build on hydra"

[Profpatsch]

Reverts #42832

We shouldn’t add hundreds of lines of copied and slightly changed source files to our package definitions. This needs to be reworked.

[Profpatsch]

@uri-canva If you want, you can open a new pull request and I can review what parts need to be changed.

[grahamc]

cc @mboes

[matthewbauer]

Sorry but there have been no PRs for alternatives yet. Bazel is a pretty important build tool and should run on macOS! Why it doesn’t is probably bazel’s fault and not Nixpkgs so I guess we should try to fix this there.

I had a branch that was a little cleaner but it never worked properly. By design bazel goes out of its way to ignore the user’s settings so some sort of hack is almost certainly needed (but i agree that it could definitely be made better than it was). But no one has done the work to do that yet.

[uri-canva]

By design bazel goes out of its way to ignore the user’s settings so some sort of hack is almost certainly needed

bazel has its own sandboxing and configuration mechanism. We can either make use of them, or hack the source to disable the sandboxing and pick up the configuration in the way we specify it in nix.

I went with the first approach, because I think it's cleaner, but it's possible to use the second approach by having two version of bazel: one with the hacks so that it can be used in nix, which can then be used to build one without hacks to be used outside of nix, skipping the compile.sh script.

[LnL7]

I don't really like that this was insta-merged without discussion/feedback from other members. Reverting a change that causes a regression is another story, but something like this which almost comes down to personal opinion should go through the same process as other nontrivial changes IMHO.

[Profpatsch]

Not to stir up an argument here, but this goes both ways; introducing such a big change (a 300% blowup in code size after all) to a package should not be merged lightly, and wait at last a few days to see if the maintainer has any comments (the PR was merged after one day).
Especially in this case, where an update PR by the maintainer was pending at the same time.

[Profpatsch]

I added a few comments to the original PR as requested.

[LnL7]

I'm not talking about the changes itself and mostly agree with the arguments there. I just think something like this shouldn't be merged instantly.