Re: [sdcc-devel] Uninited iTemp segfault / IS_LITERAL

[Sandeep D. <sa...@dd...>]

Hi Michael ,

Your fix for the == IS_LITERAL looks good.. however the crash was because
the local variable "key2" is being used before it is defined inside a loop,
the compiler
was not detecting this situation...have fixed it.. the caveat being such
variables will be spilt....

Have mode some more infrastructural changes so "make clean" required...

Sandeep

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Here's the one I found while releasing gbdk3-2.92:
>
> This code:
>
> typedef unsigned char UBYTE;
>
> UBYTE joypad(void);
>
> enum {
>    J_UP, J_DOWN, J_A
> };
>
> void main()
> {
>         UBYTE key1, key2, j;
>         j = 0;
>
>         while(1) {
>                 key1 = joypad();
>                 if(!(key1 & J_A)){
>                         if((key1 & J_UP) && !(key2 & J_UP) && j > 0)
>                                 j--;
>                    else if((key1 & J_DOWN) && !(key2 & J_DOWN) && j < 3)
>                                 j++;
>                 }
>                 key2 = key1;
>         }
> }
>
> Causes a segfault in sdcc-mcs51 and sdcc-gbz80.  If you init key2 i.e:
>
> j = 0;
> key2 = anything;
>
> it's fine.
>
> I've also made a small change to SDCCicode.c re: IS_LITERAL.  In some
> cases esp in geniCodeCast geniCodeAssign the class of an operand is
> derrived, then IS_LITERAL is run on the class.  Problem is in some cases
> when the operand is a SYMBOL then IS_LITERAL also passes, which later
> causes a segfault in floatFromVal.  I changed the check IS_LITERAL(etype)
> to op->type == VALUE && IS_LITERAL() in the cases that were segfaulting
> for me but didnt check any others...
>
> - -- Michael
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.0.0 (GNU/Linux)
> Comment: For info see http://www.gnupg.org
>
> iD8DBQE4yeLUUejL3SuzxEgRAgGbAKCgUwtC2qC8yaBqP8XtRHrgkqNCHwCgpRYz
> I+p9be/fBKJU1w7aziIfAYI=
> =OIyO
> -----END PGP SIGNATURE-----
>
> _______________________________________________
> sdcc-devel mailing list
> sdc...@li...
> http://lists.sourceforge.net/mailman/listinfo/sdcc-devel