Remove some msys2 utils

[ChrisDenton]

dos2unix should no longer be necessary. I suspect diffutils might not be needed either as (I think) only diff is used and that seems to be shipped with git bash at least.

try-job: x86_64-msvc
try-job: x86_64-msvc-ext
try-job: dist-x86_64-msvc

[rustbot]

r? @Kobzol

rustbot has assigned @Kobzol.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[ChrisDenton]

This all passed CI successfully https://github.com/rust-lang/rust/actions/runs/9487209384?pr=126343

One other thing that occurs to me is that the mingw builders use their own vendored msys2. Therefore this setup is entirely redundant. I'll try removing the msys2 action entirely and see which jobs fail. I would still expect msvc build to fail because make is currently still necessary for run-make tests.

[ChrisDenton]

Actually, thinking about it some more, I'd be a bit surprised if github CI didn't have make installed by default. So this might work for msvc even. Then again, it might have the wrong make, so we'll see.

[ChrisDenton]

Ok, that didn't work. Seems that, even for mingw builds, we do need to install msys2 make for run-make tests.

So I've reverted to the original of only adding the tools we actually need. This should be ready for now review. Oh, I forgot to mark this as draft/waiting on author... oops. So erm @Kobzol.

[Kobzol]

@bors try

[bors]

⌛ Trying commit 48d3425 with merge 818e18d...

[bors]

☀️ Try build successful - checks-actions
Build commit: 818e18d (818e18dde43bd7631b844d22b31f3755d21831ef)

[Kobzol]

The try build has succeeded, so it looks ok.

@bors r+

[bors]

📌 Commit 48d3425 has been approved by Kobzol

It is now in the queue for this repository.

[mati865]

Ok, that didn't work. Seems that, even for mingw builds, we do need to install msys2 make for run-make tests.

The errors look like win32-make was used which handles paths in a weird way. Probably can be made to work, but rewriting the tests might require less effort.
I don't think so (at least for rustc) but make might be still necessary to build C libs like openssl and so on.

[ChrisDenton]

Ok, that didn't work. Seems that, even for mingw builds, we do need to install msys2 make for run-make tests.

The errors look like win32-make was used which handles paths in a weird way. Probably can be made to work, but rewriting the tests might require less effort. I don't think so (at least for rustc) but make might be still necessary to build C libs like openssl and so on.

Ah. makes sense. I think sticking with this setup for now is the easiest to maintain. Or at least requires the fewest lines of custom bash code. It will be interesting to revisit this once the run-make tests are fully ported to rust.

[petrochenkov]

The last time I forgot to install make on a new machine or MSYS2 reinstall, I didn't notice it until trying to run run-make tests.
That's probably a good sign, unless make is involved in building something on stage 2-3, like cargo.

[Kobzol]

There is still this, for some reason 😆 I'd like to remove this Makefile finally.

[ChrisDenton]

Oh, that's weird. I didn't even know that existed. It didn't seem to fail in the jobs I checked though.

[mati865]

The last time I forgot to install make on a new machine or MSYS2 reinstall, I didn't notice it until trying to run run-make tests. That's probably a good sign, unless make is involved in building something on stage 2-3, like cargo.

I was thinking more about other tools like rust-analyzer, they have quite a dependency tree. I feel rustc is more cautious in this regard.