Hello Przemek
Przemyslaw Czerpak-2 wrote:
>
> /* compiled and linked by PellesC 4.50.15:
> * pocc.exe -Ox test.c && polink.exe t.obj
> */
>
> void * somefunc( void ) { return 0; }
>
> void * func_ok( void )
> {
> void ** gtHolder = ( void ** ) somefunc();
>
> if( gtHolder )
> return *gtHolder;
> else
> return 0;
> }
>
> void * func_gpf( void )
> {
> void ** gtHolder = ( void ** ) somefunc();
> void * pGT = 0;
>
> if( gtHolder )
> pGT = *gtHolder;
> return pGT;
> }
>
> int main( void )
> {
> func_ok();
> func_gpf();
>
> return 0;
> }
>
Here is the reply of Pelle Orinius:
Hello,
> Here is self contained C code which GPFs. OK, thanks. The -Ox option will
> enable Pentium Pro+ instructions, and enable -Ot (optimize for speed) -
> since neither -Os nor -Ot was specified. The problem is with one of the
> Pentium Pro+ instructions (conditional move), that will evaluate *all* of
> it's operands - even if they are not all used in the end, i.e. the move is
> actually performed. If one of the operands is an invalid address, it will
> GPF. Bummer.
Version 4.50 is old, so I will not make any more bug-fixes in that version.
I will look at fixing this in the current version, 5.00.
When it comes to optimizations, the -Ot option (in this case) is the much
more important than -Ox. The -Ox option will only rarely find things to
optimize, and in your case it will cause a GPF. I suggest you skip the -Ox
option and just use the -Ot option for now...
Pelle
Regards
Pritpal Bedi
--
View this message in context:
http://www.nabble.com/hb_gt_ItemBase%28%29-and-PellesC-tp22344045p22375824.html
Sent from the Harbour - Dev mailing list archive at Nabble.com.
_______________________________________________
Harbour mailing list
[email protected]
http://lists.harbour-project.org/mailman/listinfo/harbour
