Re: -W options in Makefiles

[Richard Frith-Macdonald]

On Saturday, November 9, 2002, at 12:29  am, Chris B. Vetter wrote:

> Hi,
>
> I admit I'm paranoid when it comes to compiler warnings, yet I feel 
> the following
> suggestion for GNUstep's makefiles makes some sense:
>
> I noticed that GNUstep only uses -Wall, which, in my opinion, is kind 
> of sloppy
> (sorry) since there are a lot of missing casts, plain wrong 
> parameters, missing
> and duplicated definitions and so on. Sorry again, but I think there 
> are only a
> few who wouldn't agree on that.
>
> So, I suggest to extend the inclusion of -W statements to at least the 
> following:
>
> ADDITIONAL_CPPFLAGS = -Wall -Werror -Wshadow -Wpointer-arith
> -Wcast-qual -Wcast-align -Wwrite-strings -Wconversion 
> -Wstrict-prototypes
> -Wmissing-prototypes -Wmissing-declarations -Wredundant-decls
> -Wnested-externs -Winline
>
> As I said, I'm kinda paranoid - but I'm sure the benefit is obvious - 
> cleaner code
> that actually works and doesn't b0rk out because someone passed a float
> instead of a pointer, plus a performance increase due to the fact 
> that, eg
> NSMakeRect() actually DOES use float instead of int.
>
> I'm aware that this would include a lot of work and correction, 
> however I think
> it would be worth it. At least it's something to think about...

I just spent two hours trying this with the GNUstep base library.
In that time I found no bugs.
I found  quite a few redundant declarations  which would slow compiling 
trivially.
I found cases where changing the code to avoid warnings would make it 
less readable
(pointer arithmetic and avoiding the use of the name 'index' as a loop 
index because
there is a function of that name, probably others I don't recall),
I found many cases where the warnings supplied by the compiler were 
simply wrong.

My guess is that there is reasoning behind the set of warning options 
turned on by
-Wall ... simply that the other warnings are likely to do more harm 
than good.

The result of two hours effort has been a little tidying up.  If I had 
found a single bug
I'd have considered that time well spent, but as I didn't, I'd say it 
was really a waste of
time that could have better been used working on documentation or 
regression
tests.  Perhaps a few of the additional warnings might be worthwhile.