[PATCH] avoid uninit value warnings in dead code

[Manuel López-Ibáñez]

Hi,

I think you missed my project proposal
http://gcc.gnu.org/ml/gcc/2007-03/msg00637.html

Nonetheless, your patch is interesting and I will follow closely any
follow up patches. I don't want to discourage you. I am not sure yet
how far I will get with this.

(I am not following the list closely because almost all computers I
have access to, including my own laptop that is the only computer I
own, are currently broken I am having a very very hard week, you have
no idea...)

Cheers,

Manuel.



On 12/04/07, Dirk Mueller <dmueller@suse.de> wrote:
> On Wednesday, 11. April 2007, Andrew Pinski wrote:
>
> > Actually the example Dirk gave is not a false postive "is used
> > uninitialized"
>
> It is. the variable is not used uninitialized. never. the compiler should
> realize that and not warn about something that is not true. Don't make
> useful
> warnings useless by not fixing known false positives.
>
> > In fact this was the same behavior 3.4 and before also
> > gave.
>
> Ah, so because a stone age version of gcc was wrong we have to preserve that
> behaviour? Makes no sense to me.
>
> > Let me find the old email about adding uninitialized variable
> > warning to the tree level.  It talks about this from what I remember,
> > basicially the warning is correct and actually gives us back to old
> > behavior.  Again I rather see more true postives to show up than false
> > postives to disappear.
>
> I have troubles understanding what you intend to say. If you're trying to
> say
> that in the testcase I added it should warn about "may be used" then I'm
> fine
> with that. but it is not a case of "is used" uninitialized.
>
> I can try craft a followup patch for the "maybe used uninitialized"
> behaviour
> if thats the consensus.
>
>
> Dirk
>