[cfe-commits] [Patch] Thread-safety analysis: more refactoring

[Delesley Hutchins]

> Can this happen? If this is a crash fix, there should be a testcase, otherwise
> this should be an assert.

It can happen in certain edge cases.  I'll make a test case.

> It'd be nice if this comment explained why we're doing this. As an outsider to
> this code, it wasn't immediately obvious to me that we want to extract these
> for substitution.

Done.

> What's the meaning of an invalid LockLoc? A test or an assert would elucidate.

This is me doing "defensive programming."  :-)  Some function calls --
e.g. implicit calls to destructors, don't have a location, so I don't
want to assume that the location is valid.  I'll move handling of
invalid locations to ThreadSafetyReporter.

> This comment needs updating.

Done.