Re: [PATCH 4/9] of/irq: Introduce of_irq_get()

[Thierry Reding <thierry.reding-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>]

[-- Attachment #1: Type: text/plain, Size: 1538 bytes --]

On Mon, Sep 23, 2013 at 09:18:52PM +0200, Linus Walleij wrote:
> On Tue, Sep 17, 2013 at 3:28 PM, Thierry Reding
> <thierry.reding-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> > On Mon, Sep 16, 2013 at 04:24:47PM -0500, Rob Herring wrote:
> 
> >> *_get typically also implies some reference counting which I don't
> >> believe this does. I don't think having 2 functions with completely
> >> different names doing the same thing with only a different calling
> >> convention is good either. So I would keep the old name and the names
> >> aligned.
> >
> > Okay, I'll make the new function __irq_of_parse_and_map().
> 
> I don't know why i detest __prefixing so much but I think it's
> really nasty.
> 
> Usually this is reserved for compiler- and linker related things,
> like __packed;  or __init.
> 
> I would prefer irq_of_parse_and_map_strict() or something
> like that.

Following from the discussion on one of the other patches, perhaps this
could also be converted to return int (instead of unsigned int), convert
all callers to cope with negative or 0 (instead of only 0) as errors and
then modify it to actually start returning negative error codes once all
users have been updated. Eventually I think we should get rid of using 0
as an error indicator altogether, but as I already mentioned it might be
difficult to do because new users can always come along and use the then
deprecated 0 return on error.

I'm somewhat worried about the amount of work that this is turning into.

Thierry

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]