sjlj-exceptions ICE. simplified test case

[Geert Bosch]

On Tue, 25 Apr 2000 20:49:27 -0700, Richard Henderson wrote:

  I wonder if we shouldn't work to ban sjlj for gcc 3.0.  Even
  on a.out targets we can support dwarf2 unwind tables through
  collect2.

Especially in the context of embedded systems, the space overhead of unwind 
tables may not be acceptable, so banning sjlj support is a bad thing for those 
targets. Another issue is that in some cases it might be necessary to have support
for propagating exceptions through foreign code without any tables at all.
(As somebody once said: if you find yourself wanting to remove a feature... :-)

Apart from these two concerns, there is one problem with the unwind tables
that may be fixable, or might be fixed already. At least on SGI it used to
be the case that during program initialization all addresses in unwind tables 
were patched up, which caused excessive startup delays because of swapping, even
though the actual application itself has a working set that easily fits in
memory. (This is for *huge* executables.)

  -Geert