Re: [tree-ssa] latent uninitialized variables

[Richard Henderson <rth at redhat dot com>]

On Mon, Feb 09, 2004 at 07:04:06AM +0000, Neil Booth wrote:
> >         * cppexp.c (append_digit): Rearrange unsignedp/overflow setting.
> >         (eval_token, num_binary_op, num_part_mul, num_div_op): Likewise.
> 
> Are these bugs?  Or just to suppress a bogus warning?

I'm not sure about num_part_mul, I might argue that one's a bug.

With the rest, a structure is passed (e.g. to num_trim) of which only
a couple parts are used.  Those parts are initialized, so there is no
real bug.  However, since the function isn't inlined, the compiler can't
know that.

What is true is that the variable that represents the structure member
has not been written to at the point that variable is written into the
structure temp used in the argument list.  That is what triggers the
warning.  I'm of a mind that this warning isn't wrong.



r~