[U-Boot] [PATCH 2/2] document network driver framework

[Mike Frysinger]

On Tuesday 21 July 2009 16:55:34 Ben Warren wrote:
> Mike Frysinger wrote:
> > On Tuesday 21 July 2009 03:32:55 Wolfgang Denk wrote:
> >> Mike Frysinger wrote:
> >>>> Is this a generally-accepted naming convention?  I personally think
> >>>> it's crap, and since there isn't a single driver that uses it yet, you
> >>>> might say this is a bit ahead of the curve.
> >>>
> >>> some style needed to be suggested, and what Jean proposed is better
> >>> than what we have today (which is nothing)
> >>
> >> Arent't we pretty much doing what Linux is doing here, too? I see lots
> >> of XXX_init functions in the Linux network code, for example.
> >>
> >>> that's why i said "should", deprecated current naming, and noted
> >>> existing practice.  if you agree with the proposal, it's easy enough to
> >>> run sed on a few files to fix one function name.  you agree with my
> >>> comment that today's behavior is confusing even if you stare and bang
> >>> on the code day in and day out ?  it's even worse for the occasional
> >>> observer ...
> >>
> >> Hm... renaming  something  from  "xxx_init()"  into  "xxx_register()"
> >> because  other  code  is also also using "xxx_init()" does not really
> >> make anything clearer to me. Actually IMO  it  just  adds  confusion,
> >> because  if  other's  use  "xxx_init()" I'd expect from a consistence
> >> point of view that we use "xxx_init()", too.
> >
> > your reply reinforces my point.  i'm not talking about xxx_init(), i'm
> > talking about xxx_initialize().  network drivers atm define both --
> > xxx_initialize() is to initialize the eth_driver structure and *register*
> > with the eth layer, and xxx_init() to *initialize* the hardware.  i'm
> > proposing renaming xxx_initialize() to xxx_register().
>
> I understand what you're saying, and  think in principle it's probably a
> good idea to rename to something other than xxx_initialize().  I just
> think a document that outlines best practices that are not in use *at
> all* seems a bit silly.

considering the document makes note of existing practice and suggests the new 
naming schema, i think it's fine and shouldnt hold up merging.

> If we're going to go this way, IMHO we should change all function names
> at once.  It would be easy to do, but would be a huge, potentially
> intrusive patch that I'm not sure buys us much.

how quickly we convert older drivers doesnt matter to me, but if you prefer 
sooner rather than later, that's easy enough to do and hand off to Wolfgang 
for the first patch in "next".
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20090721/8d30c793/attachment.pgp