EH bug in Makefile.in

[Mumit Khan]

hjl@lucon.org (H.J. Lu) writes:
> > 
> > hjl@lucon.org (H.J. Lu) writes:
> > > > 
> > > > I believe that the problem on linux is that you need a recent gas snaps
> hot;
> > > > 2.8.1 fills a .align in .eh_frame with NOPs instead of zeroes.
> > > > 
> > > 
> > > No. The bug is in gcc/Makefile.in. egcs-971023 turns off exceptions
> > > in C. But _eh.o in libgcc2.a has to be compiled with exceptions
> > > on. I added -fexceptions to LIBGCC2_CFLAGS in gcc/Makefile.in.
> > > EH went back to egcs-971016 for me now.
> > > 
> > 
> > I'm not so sure. Right now I have 3 egcs-971023 linux trees:
> >   
> >   a. stock: 				39 failures in total
> >   b. stock + gas-971023: 		39 failures in total
> >   c. stock + -fexceptions per HJ:	21 failures in total (but not 3!)
> >   

And the final data point:

  d. stock + gas-971023 + -fexceptions:	3 failures in total

Looks like HJ Lu's suggestion that eh.o (libgcc2.c) must be compiled with
-fexceptions is right on the money here.

Thanks
Mumit -- khan@xraylith.wisc.edu
http://www.xraylith.wisc.edu/~khan/