c++/3099: cygwin: Many g++ EH failures

[Billinghurst, David (CRTS)]

I build 3.0 with these patches, and was definitely using the shared library.
I had signal problems in gdb too, but couldn't work around them.  

I get reasonable results with --enable-sjlj-exceptions --disable-shared.
That will have to do for now, I guess.

> -----Original Message-----
> From:	Christopher G. Faylor [SMTP:cgf@redhat.com]
> Sent:	Saturday, June 16, 2001 3:35 AM
> To:	Richard Henderson; Billinghurst, David (CRTS);
> gcc-patches@gcc.gnu.org
> Subject:	Re: c++/3099: cygwin: Many g++ EH failures
> 
> On Wed, Jun 13, 2001 at 05:53:25PM -0700, Richard Henderson wrote:
> >Soliciting help from Cygwin folks:
> >
> >The following nearly works.  If g++.brendan/eh1.C is linked against
> >static libgcc, it passes.  If linked against libgcc1.dll it crashes
> >sometime after main exits -- I presume while playing with destructors.
> >
> >Most irritatingly, if I run it under the debugger, win98 crashes and
> >I have to push the reset button.  Thus the request for help -- perhaps
> >this losage doesn't show up on NT and can be debugged.  With any luck
> >it's just a matter of having built the dll incorrectly.
> 
> I just tried to build this by applying your patches to the trunk.  I
> don't have a complete 3.0 checked out.  The executable isn't using the
> shared libgcc1.dll, AFAICT, but it does crash.  A stack trace is below.
> 
> I had a hard time getting gdb to stop when the the program hit the
> abort.  gdb doesn't deal well with cygwin's signals (such as abort).
> 
> Eventually, I set a breakpoint on the first line of 'try_to_debug'.
> When it stopped there, I switched to thread 1 to get this stack dump.
> 
> I don't know if this helps or not or if I somehow built something
> incorrectly.
> 
> cgf
> 
>