formulae_detect: fetch @added_formulae too

[carlocab]

This should prevent accidentally merging PRs for new formulae without bottles.

[MikeMcQuaid]

Makes sense! When does this normally occur?

[Bo98]

Does @testing_formulae not already include @added_formulae?

[carlocab]

Makes sense! When does this normally occur?

If you press the "Merge" button instead of doing brew pr-publish, it seems: Homebrew/homebrew-core#171733

Does @testing_formulae not already include @added_formulae?

Doesn't seem so. See PR linked above.

[Bo98]

Looks like it detected correctly: https://github.com/Homebrew/homebrew-core/actions/runs/9090983591/job/24984732312#step:4:31

[carlocab]

Ah, actually @added_formulae are in @testing_formulae already. The bug is in the bottle equality check, I think.

[Bo98]

The bug is on the test rather than the detection side - we don't actually demand a bottle block exists:

homebrew-test-bot/lib/tests/bottles_fetch.rb

Lines 25 to 30 in 54dd84d

	tags = formula.bottle_specification.collector.tags
	tags.each do |tag|
	cleanup_during!(args:)
	test "brew", "fetch", "--retry", "--formulae", "--bottle-tag=#{tag}", formula_name
	end

tags in this case is probably []

[MikeMcQuaid]

Good catch @Bo98. Yeh, requiring bottles always there for homebrew/core makes sense.

(unrelatedly: I think it would be really nice if pressing the merge button did actually work one day; perhaps it could push the bottle commits to the PR and then fail?)

[carlocab]

(unrelatedly: I think it would be really nice if pressing the merge button did actually work one day; perhaps it could push the bottle commits to the PR and then fail?)

Could try calling brew pr-publish on the merge queue job

[Bo98]

I don't think merge_group has PR information, because a merge group can contain multiple merges.

[carlocab]

Yup, but each PR gets its own run within a merge_group, so might be there

[MikeMcQuaid]

Yup, but each PR gets its own run within a merge_group, so might be there

Yes, should be. Should also be possible to infer each PR programmatically.