An alias bug

[C. van Reeuwijk]

>   > For what it's worth, the latest snapshot I have (7/7/1998) gives bogus
>   > 'might be used uninitialized' messages. I'm sure they are bogus because
>   > they complain about variables that are initialized in their declaration
>   > statement. E.g.
> The code which determines whether or not a variable is initialized or
> not is very simple and gets confused by complex code.
> 
> While the initialization may be obvious in the source, certain optimizations,
> particularly gcse, may  move the initialization to a different location
> within the function, which may make it appear as if the variable is
> uninitialized to the simplistic code to find uninitialized variables.

Fair enough, but (a) I didn't get these complaints in the previous
shapshot we used (egcs-2.91.28), and (b) as a humble user I'm not happy
with these false positives, since I like my code to be warning-free, and
this forces me to abandon this warning or rewrite my code to accomodate
the analyzer.

I'm sorry I cannot offer any more constructive help here.

Sincerely,


-- 
Kees van Reeuwijk, Delft University of Technology
http://pds.twi.tudelft.nl/~reeuwijk