transpile: Convert final return statement into tail expression

[Rua]

• Fixes transpile: return is stripped from end of statement expressions #1506.
• Fixes transpile: return is emitted in final statement #1496.

[Rua]

I ran into an unexpected snag with this one: the stmts_block helper function is adding back the semicolon if it's missing. Ideally I would just remove that behaviour, but that has the obvious potential to break a lot of things, so does anyone know why it's there and what other things rely on it?

[kkysen]

It seems this is passing c2rust-testsuite CI even though there are many errors, so something went wrong there. Ugh.

[Rua]

I was able to remove the semicolon-adding code from stmts_block without any adverse effects in the tests, so let's hope it wasn't needed anywhere else. I think this is ready for review now.

[Rua]

In order to solve #1506 I had to add another variant to the ImplicitReturnType enum, because with Void it was adding a return that doesn't belong there. I don't know if that's actually the correct solution here. I don't quite understand the label break stuff in the compound statements translation, and there is likely some interaction with #1509 as well.

[kkysen]

I don't quite understand the label break stuff in the compound statements translation, and there is likely some interaction with #1509 as well.

I'm not sure if you saw, but we have #1544 now to fix #1509.

[Rua]

I don't think this PR will affect that one, but we'll have to see I suppose.

[Rua]

@ahomescu the test here is failing since merging #1544. It's inserting a label and a block, seemingly for no reason. Is that a bug in your code or in mine?

EDIT: Seems to occur with current master as well. I've updated the test so that it includes the additional block.

[ahomescu]

@ahomescu the test here is failing since merging #1544. It's inserting a label and a block, seemingly for no reason. Is that a bug in your code or in mine?

EDIT: Seems to occur with current master as well. I've updated the test so that it includes the additional block.

That's not surprising, #1544 changed a few details about how we emit the expression in a WithStmts. We should fix here whatever conflicts that caused.

[ahomescu]

There is also a clippy lint for this https://rust-lang.github.io/rust-clippy/master/index.html#needless_return. Could we just run that instead?

[Rua]

Can clippy run even when no Cargo.toml has been generated?