Re: [[gtkmm] Destroying container problem]

[Jarek Dukat <madmaxer poczta fm>]

On Fri, 2002-07-12 at 19:04, Nick Huxley wrote:
> > 
> > Nonsense, if it was a memory overrun then compiler change from 2.95 to
> > 3.1 would not solve the problem.
> 
> Not necessarily true. Often different optimisation modes can show up or
> hide pointer errors and different compilers produce different code. It
> only takes one compiler to pack some spare space here and there and an out
> by one pointer error can be made harmless.

I tested my code with -O2, -O3 and -O0 and all the time it worked (or
not worked with 2.95) the same way.

> It would be interesting to diff the machine code but probably too much
> work. gcc 3.1 is in general a better compiler though so it is possible
> that
> 2.95 had a bug in its' C++ compiler.

I'm not expert here, but Daniel mentioned something about problems with
virtual methods in 2.95. I think it may be the source of troubles.

-- 
struct Sig {
  string name("     J a r e k   D u k a t     ");
  string mail(" madmaxer (at) poczta (dot) fm ");
};


-----------------------------------------------------------------------
Kwasniewski na diecie od 4 miesiecy... >>> http://link.interia.pl/f1621