Re: [PATCH] checkpatch: add __asm__ to function-space-paren exception list

[Timur Tabi <timur@freescale.com>]

Andy Whitcroft wrote:

> It truly is not clear what "type" of thing an __asm__ is these days.
> Cirtainly it seems we can use it as an attribute of a type:
> 
>         register unsigned long __r16 __asm__("$16") = rtc_access->function;
> 
> and yet also in its more traditional form:
> 
> 	__asm__("	call foo");
> 
> The latter form feels like a function? 

But it's not.  Sure, it defines a block of code that has input and output 
parameters, but one key distinction is that __asm__ is not the name of the 
"function".

? But cirtainly in the examples it
> is shown with a space some of the time, and not others. 

Well, since checkpatch.pl insists that the space be removed, that's probably why 
it's not there a lot of the time.

> Is __asm__ an
> attribute of the null function (;) in this context or ... well ?

I don't think __asm__ can be compared to other C-language syntax constructs.

> This patch really just removes any checks for spacing on __asm__ do we
> have a preferred style for these? 

I don't know, but I do know it's wrong for checkpatch.pl to think that "__asm__" 
is the name of a function.

 > Attributes do seem to have spaces,
> though in their most attribute like usage the __asm__ "attribute" does
> not seem to be used with a space, so far anyhow.

The problem is that checkpatch.pl thinks this is okay:

	__asm__ __volatile__ ("call foo");

but it doesn't like this:

	__asm__ ("call foo");

> Oh and why are we preferring the use of __asm__ over asm?  They both
> seem valid.  Should we be recommending one over the other?

It's a toss-up.  Depending on the architecture, one version has about 60% usage 
and the other about 40%.  I thought I read somewhere that __asm__ is preferred, 
but I can't remember where I read that or who wrote it.

-- 
Timur Tabi
Linux kernel developer at Freescale