Patch for i386-darwin: test suite results

[Andrew Haley]

Andreas Tobler writes:
 > Andrew Haley wrote:
 > > Andreas Tobler writes:
 > >  > Andrew Haley wrote:
 > >  > > Sandro Tolaini writes:
 > >  > >  > 
 > >  > >  > On 19/mar/2006, at 12:21, Andrew Haley wrote:
 > >  > >  > 
 > >  > >  > > OK, so the only things now failing are the tests that rely on throwing
 > >  > >  > > exceptions from signal handlers.  Make sure that the line
 > >  > >  > >
 > >  > >  > > can_unwind_signal=no
 > >  > >  > >
 > >  > >  > > is present under i[34567]86-*-*-darwin in libjava's configure.host.
 > >  > >  > 
 > >  > >  > Yes, it's present. If I set can_unwind_signal=yes much more tests fail.
 > >  > > 
 > >  > > OK.  Well, if I were doing this I'd start with Array_3, running it
 > >  > > under gdb to see where it goes wrong.
 > >  > 
 > >  > I'd say this is missing:
 > >  > 
 > >  >   MD_FALLBACK_FRAME_STATE_FOR
 > >  > 
 > >  > see gcc/config/rs6000/darwin-fallback, darwin-unwind.h, darwin.h
 > > 
 > > Why would that matter?  It's not used becasue there is no handler
 > > installed.
 > 
 > Ok, I missed that, but simply due to the fact that a libjava port w/o 
 > signal handling is nearly useless.

It's a bit unfortunate, but not really useless.  I'd like one day to
fix the CNI layer in libjava so that a system without signal handling
is correct in all respects.

Andrew.