Re: uninitialized data in t-test-alpha.sh

[Ben Pfaff]

John Darrington <address@hidden> writes:

> On Sat, Dec 09, 2006 at 01:42:15PM -0800, Ben Pfaff wrote:
>      I'm seeing the following warning from valgrind on
>      t-test-alpha.sh:
>      
>      ==4900== Conditional jump or move depends on uninitialised value(s)
>      ==4900==    at 0x80CF84F: casefilter_variable_missing (casefilter.c:63)
>
> This highlights an issue which has been concerning me for some time.
>
> The line at question says:
>   if ( val->f == SYSMIS ) 
>
> which is fine, if the variable to which val belongs is numeric.  But
> in this case, it is a string variable (with a width less than sizeof
> (flt64) ).
>
> Can string variables ever be SYSMIS? and is it ever valid to compare
> them to SYSMIS ? 

I'd discourage going down the road of comparing strings to
SYSMIS, especially if they might be shorter than
MAX_SHORT_STRING.  You could just get unlucky and be processing a
string that happens to have the same bit pattern as SYSMIS.

I'd check the variable type before making that comparison.
-- 
"The road to hell is paved with convenient shortcuts."
--Peter da Silva