Re: Warnings in the C++ Front-End and GCC in General

[Bill Currie <bcurrie at tssc dot co dot nz>]

Gerald Pfeifer wrote:
> 
> On Wed, 9 Sep 1998, Bill Currie wrote:
> > And what about those people who insist upon compiling with -Werror (like
> > me), hmm?
> 
> Bill, do your really manage to succeed with that?

In general, yes (X11 caused some problems, but I hacked my headers to
fix them (put in int's)).  Oh, I should have been a little more clear: I
once tried to compile some page with -Wall -Werror and didn't get far
(can't remember which one).  I now (usually) only use -Wall -Werror on
my own code.

> We originally strived for a zero warning policy, but EGCS -- not our
> code! -- still has way to many bugs which trigger incorrect warnings.

I haven't used egcs all that much yet, but I *HAVE* succeeded in
compiling some my code with egcs.

> For example, as I have reported back last December, even the one-liner
>   main() { for(int i=1;;i++); for(int i=1;;i++); }
> causes
>   warning: declaration of `i' shadows previous local
> with egcs-1.1 and egcs-current when -Wshadow is used.

Doesn't -Wall trigger -Wshadow, or has that changed since gcc-2.7.2.1? 
Hmm, then again, I don't think I have any code that looks like your
main.  I think my example would look like:

   main() { if (foo) { for(int i=1;;i++); } else { for(int i=1;;i++); }}

ie the two declarations in different scopes.  However, I'm sure I've got
some code that does what your main does, but I may not have compiled it
with egcs yet. (BTW, neither of the above would compile for me: default
int, no return, etc)

I can actually see why egcs would get confused in your case: I beleive
egcs still warns about the following:

{
   for (int i=0;!data[i];i++);
   if (i) return;
}

with the `old scoping rule' warning.  I suppose this could confuse egcs
(though it shouldn't).

Bill
-- 
Leave others their otherness