Re: [tree-ssa] Unused variable warnings

[Jan Hubicka <hubicka at ucw dot cz>]

> In message <20031017193025.GO6212@kam.mff.cuni.cz>, Jan Hubicka writes:
>  >OK,
>  >I have one problem.  It is a warning:
>  >../../gcc/haifa-sched.c:2042: warning: 'index' might be used uninitialized in
>  > this function
>  >
>  >index is initialized by:
>  >if (max_issue (ready, &index) == 0)
>  > 
>  >where max_issue is function that really has path throught that really
>  >does not write into the index variable, so this warning is de-facto
>  >right, just not obvious.  What do we want to do here?
> If there's really a path were index is uninitialized, then clearly we
> want to initialize it to a safe value, both for the branch and the
> mainline.

Yes, however the path is created by inlining function max_issue and
optimizing out & and *.  So whether we realize this condition or not
depend on how much inlining we do that is somewhat ugly.

The path never actually happens in reality (max_issue either finds value
and sets it or indicate failure), but we are not smart enought to thread
the jumps.
> 
>  >Shall I simply add initializer to this variable and another one in
>  >f/lex.c as we do in usual case of this warning?
> If f/lex.c also has a real path where the variable is uninitialized, yes
> (and again, the same patch probably belongs on the mainline).

Similarly here.

Honza
> 
> jeff