Re: [gtk-list] Re: Gtk 1.2.5

[Valdis Kletnieks vt edu]

On Fri, 10 Sep 1999 20:23:27 +0200, Erik Mouw said:
> Anyway, it is not a bug in GTK+, but in the compiler. It is possible that
> this version of GTK+ *triggers* the bug in the compiler. Don't use those
> extreme optimizations, or try with a new version of the compiler
> (gcc-2.95.1 is the latest, see http://egcs.cygnus.com/).

It is also possible that there is a latent bug in GTK+, which is *exposed*
under heavy optimization.  Aliasing errors (when 2 different pointers
reference the same actual memory) are a common cause.  The code may work
at low/no optimization, because the compiler generates "stupid" code
with enough gratuitous stores and reloads to make it work, but when
the stores/reloads are optimized away, stores into a memory location
via one pointer are no longer reflected in a reference by another pointer
(because the optimizer stripped out the store because it thought it was
un-necessary).

-- 
				Valdis Kletnieks
				Computer Systems Senior Engineer
				Virginia Tech

PGP signature