Since commit 7560fa60fcdcdb0da662f6a9fad9064b554ef46c (gpio: <linux/gpio.h>
and "no GPIO support here" stubs) drivers can use GPIOs if they're available,
but don't require them.

This patch actually enables this feature, otherwise drivers will stumble
against this:

include/asm-generic/gpio.h:111: error: redefinition of 'gpio_is_valid'
include/linux/gpio.h:21: error: previous definition of 'gpio_is_valid' was here
include/asm-generic/gpio.h:121: error: redefinition of 'gpio_cansleep'
include/linux/gpio.h:60: error: previous definition of 'gpio_cansleep' was here
include/asm-generic/gpio.h:126: error: redefinition of 'gpio_get_value_cansleep'
include/linux/gpio.h:67: error: previous definition of 'gpio_get_value_cansleep' was here
include/asm-generic/gpio.h:132: error: redefinition of 'gpio_set_value_cansleep'
include/linux/gpio.h:74: error: previous definition of 'gpio_set_value_cansleep' was here

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
---
 include/asm-generic/gpio.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/include/asm-generic/gpio.h b/include/asm-generic/gpio.h
index 9c03b71..c9bcbd6 100644
--- a/include/asm-generic/gpio.h
+++ b/include/asm-generic/gpio.h
@@ -105,7 +105,7 @@ extern void __gpio_set_value(unsigned gpio, int value);
 extern int __gpio_cansleep(unsigned gpio);
 
 
-#else
+#elif defined(CONFIG_GENERIC_GPIO)
 
 static inline int gpio_is_valid(int number)
 {
-- 
1.5.5.1
--

[ message continues ]

This looks to me like pure user error ... what was anyone doing
including <asm-generic/gpio.h> in that way?  There are a *lot* of
bogus include combinations, and we don't try to "fix" them.

It should suffice to #include <linux/gpio.h> ... that causes the
definition of either (a) stubs in that file, *OR* else (b) the
platform-specific implementation of the GPIO calls, which may
(if that platform chooses) use <asm-generic/gpio.h> utilities.

NAK on this patch.

- dave
--

[ message continues ]

This is true, partially. I saw these errors for the drivers using
linux/of_gpio.h, it includes asm/gpio.h, which includes
asm-generic/gpio.h (I submitted of_gpio.h patch before linux/gpio.h work,
though git log might say otherwise). And yes, I should fix of_gpio.h too.

The thing is, I always thought that it is _good_ practice for the nested
headers to declare/define things conditionally by themselves. That way
you'll not have to fix all combinations, everything will (should) just
work. I.e.

#include <linux/of_gpio.h>
#include <asm-generic/gpio.h>
#include <asm/gpio.h>
#include <linux/gpio.h>

^^ This insanity will work.

Though with current headers, you'll easily catch the bogus users of
the asm/gpio.h thus will discipline the users, that might be a plus
too... ;-)

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

[ message continues ]