Re: [PATCH 16/16] fix handling of integer constant expressions

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

On Sun, Jun 24, 2007 at 08:18:52PM +0200, Segher Boessenkool wrote:
> >#define _IOC_TYPECHECK(t) \
> >        ((sizeof(t) == sizeof(t[1]) && \
> >          sizeof(t) < (1 << _IOC_SIZEBITS)) ? \
> >          sizeof(t) : __invalid_size_argument_for_IOC)
> >poisoning _IOW() et.al., so those who do something like
> >
> >static const char *v4l1_ioctls[] = {
> >        [_IOC_NR(VIDIOCGCAP)]       = "VIDIOCGCAP",
> >
> >run into trouble.
> 
> >The only reason that doesn't break gcc to hell and back is
> >that gcc has unfixed bugs in that area.
> 
> If I understand correctly what bugs you are talking about,
> most (all?) of those were solved in the dark ages already
> (i.e., the 3.x series).

Alas, no.  gcc is amazingly (and inconsistently) sloppy about the
things it accepts as integer constant expressions.
 
> >It certainly is not a valid C
> 
> Why not?  Nothing in the C standard says all your externs
> have to be defined in some other translation unit you link
> with AFAIK.

It's not about externs.  It's about things like

unsigned n;
int a[] = {[n - n + n - n] = 1};

And yes, gcc does eat that.  With -pedantic -std=c99, at that.  
However, 

unsigned n;
int a[] = {[n + n - n - n] = 1};

gets you error: nonconstant array index in initializer

And that's 4.1, not 3.x...