[PATCH] D47096: CodeGen: block capture shouldn't ICE

[John McCall via Phabricator via cfe-commits]

rjmccall added a comment.

In https://reviews.llvm.org/D47096#1105374, @jfb wrote:

> In https://reviews.llvm.org/D47096#1105368, @rjmccall wrote:
>
> > RecursiveASTVisitor instantiations are huge.  Can you just make the function take a Stmt and then do the first few checks if it happens to be an Expr?
>
>
> I'm not super-familiar with the code, so I might be doing something silly.
>
> I did something like this initially (leave the top of the function as-is, and instead of cast do dyn_cast to Expr and if that fails to CompoundStmt, recursively iterating all the children of the CompoundStmt). My worry was that I wasn't traversing all actual children (just CompountStmt's children), and AFAICT there's no easy way to say "take any Stmt, and visit its children if it has such a method". I could hard-code more Stmt derivatives but that seems brittle, I could use the "detection idiom" but that's silly if there's already a visitor which does The Right Thing through tablegen magic.
>
> What I can do is what I did earlier, and conservatively say it was captured if it's neither an Expr nor a CompoundStmt? Or should I special-case other things as well?


`children()` is actually defined at the `Stmt` level, and if you look at how it's implemented on e.g. `IfStmt`, you can see that it visits all of the child `Stmt`s, including the if-condition.  So it should be fine.


Repository:
  rC Clang

https://reviews.llvm.org/D47096