Type checking for contract explicit storage base location

[matheusaaguiar]

Part of #597/#15727.

• depends on Parser support for explicit storage locations #15463.

[cameel]

I have just one suggestion on how to make the error reporting less annoying. It's not critical to the PR and everything else seems to have been addressed to I'm approving regardless of that. It's simple to fix so I think we should still do that though.

There were a few minor things in tests/messages that still needed some adjustment/cleanup but I decided to just push them as fixups to avoid stalling the PR. Please take a look and squash if you accept them.

[cameel]

Looking at test output, the errors generated by this may be a bit annoying. It reports an error for every specifier in the hierarchy of every contract, which may be a lot. Consider this example:

contract A      layout at 1 {}
contract B is A layout at 2 {}

contract C1 is B {}
contract C2 is A, B {}
contract C3 is B {}

contract D1 is C1 {}
contract D2 is C2 {}
contract D3 is C3 {}

You'll get 13 error messages here. I think it would be more useful to only report an error when a contract directly inherits from one with a layout specifier. The point where a fix needs to be applied is unlikely to be far from the specifier. It's either the specifier itself being applied to the wrong contract (secondary messages already point at this) or the first contract that inherits from it (that's what the primary message should point at). This would reduce the number of messages here to 5.

This should be very easy to implement simply by iterating over baseContracts() rather than the whole annotation().linearizedBaseContract. We could then point at the exact inheritance specifier that triggered the error rather than whole contract definition (baseContracts() contains specifiers rather than contract definitions, and we can easily report their locations).

We should also have the above as a test case.

[matheusaaguiar]

Addressed in #15900.

[cameel]

As discussed on the call, we can just merge this and do #15528 (comment) as a separate PR on top.

Please remember to squash the fixups though :)

[cameel]

I see you squashed it all into one commit :) No need to go that extreme, I left it in a state that would nicely autosquash into 3 clean commits without conflicts and with test additions being done first and the effect of code changes visible separately. That's generally a structure that I'd recommend and much prefer over it being smooshed together. Can you restore that version?