lily on ia64 (cont'd)

[Han-Wen Nienhuys]

address@hidden writes:
> Hi,
> 
> FYI: on ia64, with gcc 3.3.1 and bison 1.28, lily compiles and seems to
> run (though I still have problems with scm load path, font path, etc., but
> that is another story).  Putting debugging code into the destructor of
> Array however shows, that many times array_ is already 0 when entering the
> destructor.  This is even true on i586, but much more seldom than on ia64.
> With gcc 2.96, lily crashed with segfault in the destructor.  A simple
> test program on ia64 with gcc 2.96 that applies the delete operator on a
> null pointer, however, did *not* immediately crash; there must be
> something additional that causes the crash.
> 
> I think it should be considered a bug that array_ sometimes is 0 when
> entering the destructor (including i586), even if it does not directly
> seem to immediately crash lily, right?

Delete of a null pointer has no side-effects. That's standard C++,
AFAIK, and I have never tried to prevent it.

--
Han-Wen Nienhuys   |   address@hidden   |   http://www.xs4all.nl/~hanwen