Re: new improved EH on Darwin

[Geoffrey Keating <geoffk at apple dot com>]

On Monday, November 18, 2002, at 03:24  PM, Richard Henderson wrote:

> On Wed, Nov 06, 2002 at 01:00:53PM -0800, Geoffrey Keating wrote:
> > 	* g++.old-deja/g++.eh/badalloc1.C: XFAIL excess errors test on
> > 	Darwin.
> If this passes, then we're not testing what we think we're testing.
> The C++ runtime _does_ allocate memory when throwing.  There is no
> way you could have affected that.  This should still fail.
This is the 'excess errors' test; it fails on Darwin because this test 
overrides malloc(), which causes the linker to complain about duplicate 
symbols.

However, the runtime test does pass for me (annoyingly, XPASS because 
the test harness thinks the excess errors mean that compilation failed) 
on Darwin.  When I look at it, the runtime seems to always have a 
fallback for being unable to allocate memory, for instance, see 
emergency_buffer in __cxa_allocate_exception.