On Tue, Mar 12, 2013 at 11:39 AM, Adrian Prantl <[email protected]> wrote:

>
> On Mar 6, 2013, at 1:15 PM, Eric Christopher <[email protected]> wrote:
> > Sorry for the delay, I'm wondering if there's a way to look at the
> location in the backend - even without the column information to make a
> determination here. I'll probably look into it before putting this in
> though.
>
> To summarize the problem:
> LexicalScopes::getOrCreateInlinedScope() uses the DebugLoc as an index
> into InlinedLexicalScopeMap.
> DebugLoc::getFromDILocation(MDNode *N) builds a new DebugLoc from
> LineNumber, ColumnNumber, Scope, and OrigLocation of InlinedAt. In my test
> case, Line, Scope and OrigLocation are all identical for two inlined calls
> to the same function at the same line.
>
>
This is what I was hoping wasn't the case. :(


> Based on this I think (but I may be wrong) that we need to emit a distinct
> location already in the frontend (in CodeGenFunction::EmitCallExpr). The
> backend relies on the DebugLoc to create the lexical scope.
>
> Do you see another way that you would prefer?
>

Unfortunately not at the moment. I guess go ahead then.

Thanks for looking into it.

-eric
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to