compiletest: Remove the libtest-based executor and its dependency

[Zalathar]

Now that #140288 has landed and the new compiletest executor is used by default, we can now move forward with removing the libtest dependency from compiletest.

My hope is that after landing this, we can configure bootstrap to build compiletest with the pre-built stage0 library by default, instead of the in-tree stage0 library. That would give the stage0 redesign one less thing to worry about.

---

This PR has deliberately been kept small and simple, to make it easier to revert if necessary. Further cleanup can take palce after we're confident that it won't need to be reverted.

r? jieyouxu

Blocker for #119899

[rustbot]

Some changes occurred in src/tools/compiletest

cc @jieyouxu

[jieyouxu]

I think we should keep the libtest executor until the next stable release (2025-05-15 (in 16 days)) in case something goes horribly wrong (seems very unlikely, but as a backup plan) during stable release or whatever. Wdyt?

EDIT: but I guess we'd have to do changes and reverts anyway.

[Zalathar]

Hmm, at the very least it’s probably a good idea to avoid any subsequent cleanup that would make this harder to revert, until after the next beta.

But I think it should be fine to land something like this change (or perhaps an even easier-to-revert variant) in the near-ish future.

[jieyouxu]

If we really ought to

avoid any subsequent cleanup that would make this harder to revert, until after the next beta

then I think we might want to hold this off until next beta anyway. We can still land stage 0 redesign PR without this PR (if that might happen), keeping the compiletest-stage0 hack, and then rip out the stage0 hack after landing stage 0 redesign, then this.

[Zalathar]

In the medium-term (i.e. after this and the stage 0 redesign have both landed), the situation I hope to be in is:

• Compiletest uses pre-built stage0 std by default for all contributors (despite using internal_output_capture).
• There is a bootstrap flag to force it to use in-tree stage1 std instead, but this is only enabled in a small number of CI jobs (to make sure it actually works), for the benefit of distros etc that want to bootstrap with a same-version compiler instead of previous-beta.
• In the unlikely (?) event that the internal_output_capture API changes before we eventually stop using it, compiletest would use cfg(bootstrap) to account for the differences.

Removing the libtest dependency before landing the stage 0 redesign is not strictly necessary, because we do have the stage0 hack, and it's likely that the libtest dependency will be removed before any internal details of libtest actually change (which is what would break the hack).

But I would prefer to not to force compiler contributors to have to think about the hack if reasonably possible.

[jieyouxu]

I thought about this some more. Let's land this, I'll put a hold on any further compiletest adjustments that would complicate reverting libtest removal until next beta. Agreed that ideally, compiler contributors won't have to worry about the compiletest stage0 hack.

@bors r+ rollup

[bors]

📌 Commit cbfc364 has been approved by jieyouxu

It is now in the queue for this repository.

[jyn514]

this is really amazing work, thank you both so much <3