No uninitalized report in a pre-returned match arm

[linyihai]

This is an attemp to address #126133

[rustbot]

r? @pnkfelix

rustbot has assigned @pnkfelix.
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

[pnkfelix]

This solution does not cover all the cases I can imagine, but it is an improvement over the prior overall behavior, so I'm going to r+ it.

In particular, special casing return statements does not cover other kinds of non-local control-flow operations, such as break or continue. (But, to be clear:you cannot use a simple walk over the AST to identify which break/continue will cause the match arm to skip over the relevant use of the potentially uninitialized variable.)

The big picture problem here is that precisely identifying which match arms both A. can reach a use of the variable and B. fail to initialize the variable requires a control-flow analysis. Which you can get from the MIR, but you're trying to report a diagnostic with respect to the HIR.

(My suspicion is that a fully precise solution would work by using the MIR to identify the relevant spans that meet condition A and B in the previous paragraph, and then seek out the match arms here that overlap such spans. Or at least, that would be how I might investigate trying to attack a fully precise solution. But there are probably other rustc developers with alternative approaches in mind.)

[pnkfelix]

@bors r+

[bors]

📌 Commit 5d8f40a has been approved by pnkfelix

It is now in the queue for this repository.

[pnkfelix]

@bors rollup