Re: GNOME CVS: gnome-core martin

[Maciej Stachowiak <mjs noisehavoc org>]

On 12Aug2001 10:45PM (-0400), Owen Taylor wrote:
> 
> The question is not really whether warnings stick around, but
> whether bugs stick around.
> 

In my experience at least 1/3 of warnings turn out to be actual bugs,
though sometimes obscure ones. 

> -Werror:
> 
>  - Is somewhat of a wash for code quality. If you carefully review
>    every warning, and carefully review your fixes, you probably
>    make a net improvement. But careless squashing of warnings
>    can easily introduce bugs, or put casts in the wrong place
>    preventing the right fix from being made.

Hmm, I have seen fixing warnings fix bugs, but I've never seen a
confirmed sighting of it inroducing a bug.
 
> People are, of course, free to use -Werror in their modules, and
> are free to put -Werror in their CFLAGS, and send us patches.
> 
> Warning suppression takes careful attention, and context switching
> to do that when a build dies is expensive, even if you know
> the code sufficiently well.

The best time to fix a warning is as soon as possible after you
introduce it, much as with any other bug. This is why -Werror being on
by default is a good thing - it prevents warnings from ever getting in
the state where a context switch is needed to fix it.

But anyway, it's a matter of taste, and not worth debating.

 - Maciej