Re: [RFC PATCH] cpumask_t initializers

[Janis Johnson <janis187@us.ibm.com>]

On Fri, Nov 19, 2004 at 05:45:57PM -0800, Matthew Dobson wrote:
> 
> Janis Johnson, a GCC hacker, told me the following:
>> Extra parens can be thrown away in expressions, but the syntax for
>> initializers has curly braces on the outside of the list.  GCC doesn't
>> seem to mind if there are parens outside the braces for a struct
>> initializer, but that's probably a bug and could change in a future
>> version of GCC's C parser.
> 
> So, in order to both make my code compile and future-proof (heh) the
> CPU_MASK_* initializers I wrote up this little patch.  This DEFINITELY
> needs to be tested further (I've compile-tested it on x86, x86 NUMA,
> x86_64 & ppc64), but the good news is that any breakage from the patch
> will be compile-time breakage and should be obvious.
> 
> The fact that GCC's parser may change in the future to disallow struct
> initializers wrapped in parens kind of scares me, because just about
> every struct initializer I've ever seen in the kernel is wrapped in
> parens!!  This needs to be delved into further, but I'm leaving for home
> for a week for Thanksgiving and will have limited access to email.

I'm not an expert on the C language (I just pass for one in this
building) so this ought to be looked at by someone who is.  As for
changes to the C parser, Joseph Myers is writing a recursive-descent
C parser for GCC tentatively slated to replace the existing C parser
for GCC 4.1.

Janis