linux/diagnostic: add check for versioned GCC linkage

[carlocab]

• Have you followed the guidelines in our Contributing document?
• Have you checked to ensure there aren't other open Pull Requests for the same change?
• Have you added an explanation of what your changes do and why you'd like us to include them?
• Have you written new tests for your changes? Here's an example.
• Have you successfully run brew style with your changes locally?
• Have you successfully run brew typecheck with your changes locally?
• Have you successfully run brew tests with your changes locally?

---

This complements my other two GCC-on-Linux PRs (#13631, #13633), however
they are both reliant on bottles eventually being (re-)poured.

Let's try to speed that up by returning an error message from brew doctor
whenever a user has formulae installed that would benefit from a brew reinstall.

[BrewTestBot]

Review period will end on 2022-08-04 at 13:00:15 UTC.

[carlocab]

As before, I still need to test this. But this is the proof-of-concept.

[carlocab]

This could be very slow. On the other hand, brew doctor isn't run all that often.

[MikeMcQuaid]

@carlocab Would be good to benchmark how much slower it ends up being?

[carlocab]

@Homebrew/linux would appreciate help with testing and benchmarking. Using Docker on macOS is not fun 😅

[carlocab]

Ok, this may take me a bit, but the exercise I plan to do is to install 100 randomly selected GCC dependents:

brew ruby <<EOS | xargs brew install
require 'formula'
gcc_dependents = Formula.all.select { |f| f.deps.map(&:name).include? "gcc" }
puts gcc_dependents.sample(100)
EOS

and then test brew doctor against this branch and against master.

[carlocab]

Assuming I did this correctly, brew doctor benchmarks should be in the markdown summary on this page in about an hour or so: https://github.com/carlocab/workflow-test/actions/runs/2804210379

[carlocab]

System.IO.IOException: No space left on device : '/home/runner/runners/2.294.0/_diag/Worker_20220805-142208-utc.log'

Oops.

[carlocab]

What I should be doing here is checking that the current directory precedes the versioned directory whenever the versioned directory is found, but I'm still thinking of the right way to do that.

In any case, if this check is something we want to do, then this is still an improvement over not checking, and we can iterate on this later. But, suggestions very much welcome.

[MikeMcQuaid]

Approach seems reasonable. Perhaps worth doing a benchmark and maybe checking on the setting that should have migrated all of these?

[Bo98]

Is this for the scenario where the user doesn't use brew update?

[carlocab]

I'll admit to not having thought too carefully about failure modes yet: my motivation is that it would be good to provide as many paths to brew reinstall as possible to minimise breakage.

Something like this (omitting the #core_tap? check) might help for your concern over source-built formulae from #13633 (comment), though.

[BrewTestBot]

Review period ended.

[carlocab]

Benchmarked:

Before

Command	Mean [s]	Min [s]	Max [s]	Relative
brew doctor	5.248 ± 0.053	5.166	5.328	1.00

After

Command	Mean [s]	Min [s]	Max [s]	Relative
brew doctor	2.915 ± 0.033	2.862	2.957	1.00

This is on a system with about ~430 formulae installed, which is pretty extreme.

This means brew doctor is nearly twice as slow as before, which is pretty bad. OTOH, it didn't take long at all to run before, so it's not too bad in the end.

[MikeMcQuaid]

This means brew doctor is nearly twice as slow as before, which is pretty bad. OTOH, it didn't take long at all to run before, so it's not too bad in the end.

Yeh, I think that's too slow to run for everyone every time 😭.

[carlocab]

Running some slightly more comprehensive benchmarks here: https://github.com/carlocab/workflow-test/actions/runs/2805207136/attempts/1#summary-7696068383

One of the runs is still going, but the ones that finished are a bit all over the place.

I'm ok to leave this be for now. We can reconsider it depending on the number of users who report issues that might be helped by this.

[MikeMcQuaid]

I'm ok to leave this be for now. We can reconsider it depending on the number of users who report issues that might be helped by this.

Shouldn't this be handled automatically by brew update? If so, can we check if the setting is set, make this a no-op if so and tell them to run brew update if anything fails? Similarly, we could make it fail the first time we find a single failure rather than getting all the failures from all formulae.

[MikeMcQuaid]

Let's ship this and revert if it's causing issues/too slow.