#389 M68K/BCPL: compstring (a.k.a. stricmp) problem

[Mark K]

The BCPL compstring function (called stricmp in the AROS source) at GV offset $134 has a problem.

The string comparison clears bit 5 of the bytes read from each string. It should only clear bit 5 if the character is between 'a'-'z'. [The AROS versions of compch (charicmp in the AROS source) and capitalch (toupper) already do that correctly.]

Also, I think the ext.l d0 instruction at the end could be removed if you change the sub.w d1,d0 to sub.l d1,d0.