Use output directory naming scheme diff_against_dynamic_baseline

[fmeum]

Compared to diff_against_baseline, this mode improves caching when top-level flags change that are reset in the exec configuration.

Fixes #18480

[fmeum]

@linzhp Could you test the change? I did and it seemed to resolve the issue.

[linzhp]

It worked. Thanks!

[gregestren]

@sdtwigg for triage - do we still need this PR?

[sdtwigg]

Okay, it has been about a month since my change was submitted so users have had a chance to test and reported no issues (at least compared to the old diff_against_baseline).

I think it might be reasonable to consider setting this as the new default. Any thoughts?

PS: What is with the objc thing? Can you use the helper-function in CoreOptions instead of the raw equality check?

[aiuto]

Do you need a global blazerc change before this? We can discuss in the import

[fmeum]

PS: What is with the objc thing? Can you use the helper-function in CoreOptions instead of the raw equality check?

Thanks for the pointer, I wasn't aware of the helper function. Changed.

Do you need a global blazerc change before this? We can discuss in the import

As far as I know google3 already uses a different value than Bazel. If that's still the case this wouldn't be needed.

[gregestren]

Do you need a global blazerc change before this? We can discuss in the import

As far as I know google3 already uses a different value than Bazel. If that's still the case this wouldn't be needed.

This is correct.

[fmeum]

@sdtwigg Friendly ping, it would be great if we could get this into a rolling release soon so that it can enjoy testing "in the wild" well before the 7.0.0 branch cut.