Libjava failures status

Sun Mar 10 04:59:00 GMT 2002

> On Thu, 2002年03月07日 at 12:28, Joel Sherrill wrote:
> > 
> > 
> > Jan Hubicka wrote:
> > > 
> > > Hi,
> > > The yesterday libjava failures are caused by my sanity checking I added
> > > to verify_flow_info. I must say I don't understand how the patch has regtested
> > > for me before I commited it, perhaps I disabled checking :(
> > > I've fixed all the failures that happends on i386/PPC compilation today,
> > > all of them latent bugs.
> > 
> > I haven't had a chance to dig through all the 39 cross builds I
> > ran overnight for failures, but the mcore-elf and sh-* failed 
> > building newlib's strftime.c with an ICE in verify_local_live_at_start 
> > at flow.c:591. Is this somehow related?
> > 
> This may or may not be related, but the overnight SPEC95 runs failed
> with a similar failure while building 134.perl:
>> /home/dnovillo/perf/sandbox-gcc/i686/inst/bin/gcc -c -o form.o -O2 form.c
> Error from 'specmake build 2> make.err > make.out':
> form.c: In function `format':
> form.c:353: Internal compiler error in verify_local_live_at_start, at flow.c:607

Debugging the problem points me to combine liveness updating. I've sent
a fix for it day before yesterday and I guess it is the same problem as you
are encontering. Since Spec95 seems to work now I am hoping it is fixed.
Let me know it case it is not.
Does the mcore problem remain? I will check myself at tuesday
but knowing that it works now would save me some time :)
Honza