[cfe-commits] r139437 - in /cfe/trunk: include/clang/AST/Stmt.h lib/AST/Stmt.cpp lib/Analysis/CFG.cpp test/SemaCXX/return-noreturn.cpp

[Chandler Carruth]

On Fri, Sep 9, 2011 at 9:49 PM, Ted Kremenek <kremenek at apple.com> wrote:

> Ah, nevermind.  I misread the patch.


Whew! Ok. Yea, I looked at that, and didn't intend to change that. Just to
make sure I *did* understand this correctly, my impression was that this is
controlling how we build up the parent block, prior to any splitting done by
the introduction of cases, defaults, or labels. We need to see all the
destructors and other delayed-effect inducing constructs.

That said, maybe we could do this when we visit the various statements?
Essentially try to do the DeclStmt processing at the start of each of those
visit methods? It would duplicate the logic here, but would result in only
walking the chains of these once. Thoughts?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20110909/b7c27147/attachment.html>