Re: [PATCH] Tidy up 'make' output

[Pavel Roskin <proski@gnu.org>]

On Wed, 2008-06-18 at 19:46 +0200, Robert Millan wrote:
> On Tue, Jun 17, 2008 at 03:31:54PM -0700, Colin D Bennett wrote:
> > I'm all for warning-free code, but if we try to
> > use -Werror, the code won't even begin to compile in the current state.
> 
> Of course, I wasn't proposing to add -Werror in the current state and just
> throw the hot potato into everyone ;-)
> 
> Ideally, someone (or all of us ;-)) could do the work to eliminate those
> warnings, then add -Werror, and at that point it's the responsibility of
> every contributor that new code is warning-free.

There will be some combinations of gcc and libraries that will produce
warnings.  It should be easy to turn off -Werror on the make command
line if necessary.

> So is the proposed situation you don't like, or the path that would be
> needed to archieve it?

That's OK, but it's doesn't make build system changes unnecessary.  The
less noisy build system will help find other messages that -Werror won't
catch, such as linker warnings.  It will help understand what is
happening and what is potentially wrong or suboptimal.

For example, I'm seeing warnings from xfs.c that nobody is fixing.  I
can fix the warning by changing the code so that it does exactly what
it's doing now but doesn't cause a warning.  The problem is, I don't
see corresponding structures in the Linux xfs code.  I don't have time
to investigate xfs implementation to see if I'm possibly hiding a bug.

I can imagine that somebody knows more about xfs but doesn't see the
warning.  Once we make the warning visible, maybe that person will have
a look and make a better patch.

It's also possible that somebody who want to install GRUB in xfs will be
extra cautious when seeing the warning.  It's actually a good thing.
Sure, not having the warning will be even better.

-- 
Regards,
Pavel Roskin