Re: exceptions: gross hideous hack works, need guru to tell me ifI have caused major breakage

[Andrew Haley <aph at cambridge dot redhat dot com>]

Jeff Sturm writes:
 > On Sat, 15 Dec 2001, Andrew Haley wrote:
 > > This logic has been moved from Java's own signal handler to
 > > MD_FALLBACK_FRAME_STATE_FOR in the various gcc back ends.  There are
 > > versions for x86, ppc, Alpha and recently IA-64.
 > 
 > On GNU/Linux, that is.

Ah yes, I should have mentioned that.

 > >  > 3) What I've done works, but yields invalid stack traces (which don't
 > >  >    work on Win32 anyways due to SJLJ, so it's all moot).
 > > 
 > > Well, that's a shame, but only to be expected.

 > Hmm?  I don't follow... getting a stack trace has nothing to do with the
 > unwinder.  On x86 it is straightforward.  The last frame pointer
 > is always stored at (%ebp) and the return IP at 4(%ebp).

I don't understand why stack traces in general shouldn't work on
Windows.

 > (I don't know if you can step through a win32 exception handler
 > frame this way, but alternatively you could take the faulting
 > eip/ebp and walk the stack from there.)

That's right.  This is the problem about which I said "only to be
expected."

Andrew.