Re: [PATCH v2] of_gpio: Return GPIO flags from of_get_gpio()

[Anton Vorontsov <avorontsov@ru.mvista.com>]

On Wed, Nov 26, 2008 at 02:35:54PM -0800, Trent Piepho wrote:
> On Thu, 27 Nov 2008, Paul Mackerras wrote:
> > Anton Vorontsov writes:
> >
> >> Can we apply it? Paul, Benjamin?
> >>
> >> The patchwork url for this patch is:
> >> http://patchwork.ozlabs.org/patch/6650/
> >>
> >>
> >> Thanks!
> >>
> >>>  drivers/mtd/nand/fsl_upm.c              |    2 +-
> >>>  drivers/net/fs_enet/fs_enet-main.c      |    2 +-
> >>>  drivers/net/phy/mdio-ofgpio.c           |    4 ++--
> >>>  drivers/of/gpio.c                       |   13 ++++++++++---
> >>>  drivers/serial/cpm_uart/cpm_uart_core.c |    2 +-
> >>>  include/linux/of_gpio.h                 |   21 +++++++++++++++++----
> >>>  6 files changed, 32 insertions(+), 12 deletions(-)
> >
> > That would need acks from Jeff Garzik and David Woodhouse.
> >
> > Alternatively you could add a new function (called, for instance,
> > of_get_gpio_flags) with the extra parameter to eliminate the need to
> > change any drivers at this stage, since they all seem to pass NULL for
> > the flags argument.
> 
> But if we did this every time any exported function needs to change, think
> how bloated the API would be with cruft.

Stable API is nonsense, yes. But we tend to change the API evolutionary,
not revolutionary. That is,

1. Implement of_get_gpio_flags();
2. Now we can start using it (no stall in development, see?);
3. Then somebody comes with the _cleanup_ patch:
   "[PATCH] Merge of_get_gpio_flags() and of_get_gpio(), convert users"

   ^^ That patch is trivial and could be applied at any appropriate
   moment (i.e. when there are no of_*_gpio*() patches queued in the
   -next trees). And as time goes by, the patch collects all the needed
   Acks, no need to hurry -- it's trivial cleanup.

-- 
Anton Vorontsov
email: cbouatmailru@gmail.com
irc://irc.freenode.net/bd2