-O2 failure with new ABI patches

Tue Jan 16 06:59:00 GMT 2001

Bryce McKinlay wrote:
> IIRC, it seems to have something to do with the static allocated arrays.

Yes. Somehow -O2 drops the entire array initializer, causing an unreferenced
symbol.
> Are you sure this is a new ABI regression?

Either that or some recent patch. It definitely wasn't happening with my
previous gcc build. Also I had been building libjava using -O2 -fno-defer-pop
without incident.
> It goes away if you give the compiler class files to compile against (ie still
> compiling from .java source, but also having .class versions which the
> compiler uses for dependencies).

Hmm... I'm not seeing that. I'd expect it to go away when compiling bytecodes
however, since static arrays are then not possible.
--
Jeff Sturm
jeff.sturm@commerceone.com