[cfe-dev] r161350 breaks the intention of ParentMap::addStmt

[Jordan Rose]

Hi, Olaf. I've gone ahead and just dealt with PseudoObjectExpr and BinaryConditionalOperator explicitly in r165355. I was imagining the case where some day you or someone else is using ParentMap the way you're using it -- to keep the parents of a tree of statements up-to-date, as is specifically mentioned in the header comment -- and would need it to work correctly even for OpaqueValueExprs. I've removed all checks for an existing parent, and while things look quite a bit more complicated, I don't think it will actually change the performance characteristics in any significant way.

Essentially, OpaqueValueExprs are now "transparent" by default, but PseudoObjectExpr and BinaryConditionalOperator deliberately switch to an "opaque" mode when handling their semantic sub-expressions.

Let me know if you're still having problems, and sorry about messing this up in the first place.
Jordan


On Oct 5, 2012, at 6:52 , Olaf Krzikalla <Olaf.Krzikalla at tu-dresden.de> wrote:

> Hi Jordan, hi John,
> 
> I've posted another patch at Oct.2 regarding the issue. Is there any hope that we will get a version fulfilling yours and mine requirements?
> 
> Best regards and sorry for being impatient
> Olaf