IA-64 non-call exceptions

Fri Jun 22 05:55:00 GMT 2001

Richard Henderson writes:
 > On Tue, Jun 19, 2001 at 04:22:41PM +0100, Andrew Haley wrote:
 > > The biggest issue is that (unlike the DWARF 2 unwinder) the IA-64
 > > unwinder can't cope with unwinding leaf functions. There are several
 > > reasons for this, but one is that such functions don't save a copy of
 > > the return link but leave it in br0; the unwinder can't cope with
 > > that. In any case, the compiler/assembler outputs no unwind
 > > information for leaf functions so the unwinder aborts.
 > 
 > Hmm. I _think_ we're supposed to assume that in the absense of
 > unwind information that b0 contains the return address, and that
 > there is no register frame or stack frame. I'll have to look at
 > the abi again...
Yeah, but 'twould be nice in that case at least to emit a basic
unwinder frame. As it is we emit zilch.
Andrew.