Re: [Qemu-devel] [PATCH v2] Add infrastructure for QIDL-based device serialization

[Michael Roth <mdroth@linux.vnet.ibm.com>]

On Wed, Sep 26, 2012 at 12:33:17PM +0200, Paolo Bonzini wrote:
> Il 26/09/2012 12:20, Kevin Wolf ha scritto:
> >>> >>     QIDL_DECLARE(RTCState) {                                                                            
> >>> >>         ISADevice dev qidl(immutable);
> >>> >>         MemoryRegion io qidl(immutable);
> >> > 
> >> > Just like sparse is a "compiler", so is qidl.  We are free to use the
> >> > '_' + lowercase prefix.
> >> > 
> >> >           ISADevice _immutable dev;
> >> > 
> >> > It's an established practice in wide-use.
> > Not commenting on the underscore, but you did one thing that I want to
> > support: Put the (q)_immutable in a place where it looks like a
> > qualifier. Not so important for the qidl(...) syntax, but with the
> > simplified forms I definitely like it better.
> > 
> > I think I would even have made it '(q)_immutable ISADevice dev;', but
> > having the field name last is what really matters for readability.
> 
> Agreed.  I don't want to be a nuisance, so: Michael, please pick one between

Not a problem, the parser supports both before/after. I prefer before as well,
except in the case of q_property("name", <options>) where we often need to put
the variable name on a second line, but those aren't too common so let's just
standardize on before for now since that'll benefit the common use case better.

> 
>     ISADevice QIDL(immutable) dev
>     ISADevice q_immutable dev
>     ISADevice qidl(immutable) dev
> 
> and if you choose the second, let's make QIDL an implementation detail,
> i.e. document that every new attribute we introduce should define a new
> q_* macro.

Ok, sounds like a plan. let's do q_*.

> 
> Paolo
>