This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: mainline "exploding"


> At Fri, 17 Oct 2003 09:27:25 +0200, Jan Hubicka wrote:
> > > (unit-at-a-time causes glibc to lose in other ways, too, on certain
> > > platforms at least, but that's not unit-at-a-time's fault.  See
> > > http://sources.redhat.com/ml/libc-alpha/2003-10/msg00013.html for
> > > amusement.)
> > 
> > There is problem with renaming the symbols using asm("...").  The review
> > of this patch has sucked because we are not able to decide on whether we
> > want to support all the funny side cases old syntax has or restrict it
> > in some sane way.  But this will get fixed until official release too
> > for sure.  I will try to re-open this thread.
> 
> IMO, that glibc code is sufficiently ... special that i'm surprised it
> ever worked at all.
> 
> My take on it would be, make them compile it with -fno-unit-at-a-time.
> In fact, that seems like the perfect solution, precisely because that
> file really *is* depending non-unit-at-a-time behaviour...

We are discussing possibilities to drop -fno-unit-at-a-time mode from
GCC at least in longer term, so we should get this solved.
hammer-branch GCC is able to build glibc in unit-at-a-time with
exception of one file that postprocess assembly output of the compiler
so it should not be too hard to get mainline into similar shape.

Honza
> 
> 
> Thanks for looking at the ICE, BTW.
> 
> 
> 
> chris


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]