Linus Torvalds wrote:
> On Mon, 12 Feb 2007, Christopher Li wrote:
>> On Mon, Feb 12, 2007 at 10:28:19AM -0800, Linus Torvalds wrote:
>>> (It is perfectly proper to have a typedef that is actually of a function 
>>> type, so this looks like a sparse bug regardless, it looks just as if we 
>>> don't turn a function type into a function pointer type when we see it as 
>>> an argument in the declaration).
>> Yes, we does, in examine_fn_arguments(). But not correctly inherent the attribute bits.
> 
> Ahh. 
> 
>> I think this patch should fix it, I haven't try it myself on this bug yet.
>> It works on a different test case "function vs function ptr".
>> It has been posted to the list before. It is in my series as well.
> 
> This looks good. Ack.  Josh?

See my response to Christopher later in the thread; his revised patch appears
to solve the problem fully.

> The only thing that I reacted to is that maybe we should change the
> 
> 	s->ctype.modifiers = 0;
> 
> a few lines down into a
> 
> 	s->ctype.modifiers &= ~MOD_PTRINHERIT;
> 
> or something?

Christopher's revised patch does exactly that.

- Josh Triplett