Re: [PATCH 7/7] asm-generic: suppress sparse warning in ioctl.h

[Al Viro <viro@ZenIV.linux.org.uk>]

On Wed, Apr 02, 2008 at 05:57:54PM -0700, Linus Torvalds wrote:
> 
> 
> On Wed, 2 Apr 2008, Harvey Harrison wrote:
> >
> > 1 ? 0 : x
> > 
> > is not valid in contexts where C requires integer constant expressions.
> > Index in static array initializer is one of those.
> 
> So I don't much like this one, because (a) we could just make sparse 
> accept it and (b) gcc _does_ accept it and gives us nicer error messages. 

Umm...  How far do you want sparse to go?  You _really_ don't want
bug-for-bug compatibility with gcc - it's far too weird (and that's
even before going into the effects of optimization flags).

BTW, what happened with sparse.git?  The last changeset in there (in
/pub/scm/devel/sparse/sparse.git/ on g.k.o) is
commit a02aeb329d5a8f9047c0b75b7e7f64ee2db3ffcf
Author: Josh Triplett <josh@freedesktop.org>
Date:   Tue Nov 13 04:15:13 2007 -0800

    Makefile: VERSION=0.4.1

and I definitely had seen patches on sparse maillist since then (hell,
sent several myself - including fixes for show_typename(), etc.)

I don't mind doing more liberal ICE handling, *if* we agree on a well-defined
extensions to what C99 says.  But I'd rather have some idea of what's pending
the inclusion into the tree...

As for the extensions...  Amend 6.6p6 in a way similar to 6.6p3 (i.e. allow
any junk in unevaluated subexpressions)?  Making that option-controlled,
probably...

BTW, gcc is very definitely buggy - int a[1 + 0 * x]; is accepted and
that breaks even 6.6p3, let alone 6.6p6.  With -pedantic -std=c99 -Wall,
at that.