eliminating debug info for symbols/types in unexpanded inlines

[Jim Wilson]

On Thu, 2009-05-07 at 16:39 -0700, Cary Coutant wrote:
> Offhand, though, I don't see any reason why what you're doing in this
> patch would interfere in any way with comdat type info. It looks to me
> like your patch is eliminating debug info that isn't at all needed,
> while mine is only eliminating duplicate info. It certainly looks
> worthwhile to me.

The stabs duplicate header file elimination stuff works best if you
always gets the same (full) debug info for headers.  If you get
different debug info each time you include a header file, it won't work.

I noticed your stuff, but haven't looked at it too closely.  If you are
doing elimination at a lower level, then there won't be a conflict.  If
each type gets its own comdat section for instance, then it won't matter
if the type info is only emitted by some includes and not others.  It
just means fewer type duplicates to eliminate at link time.  OK, so that
is one less thing for me to worry about.

Jim