Hi Marcel,
On 10 October 2014 16:44, Marcel Ziswiler marcel@ziswiler.com wrote:
Hi Simon
On Fri, 2014-10-10 at 16:22 -0600, Simon Glass wrote:
Well you fixed that bug. Are there any others?
Well, not in any of our boards but a short grep through the sources reveals dozens of places where GPIOs are still reserved with NULL labels. Happy crashing and subsequent bisecting for all them folks I guess.
Ah OK. It sounds like people know that the driver ignores the name and so NULL is OK, and that is no longer true. The function signature for gpio_request() just says:
* @param label User label for this GPIO
with no indication that NULL is OK. So I did not see this as a semantic change. But we have to deal with reality.
Checking for obviously invalid args is not typically done due to the code overhead (e.g try to use assert() which is compiled out in production code).
Agreed but rather sad if the semantics certainly changes.
Anyway, I give in. Let's just drop it then.
I think we need to have this patch due to the existing code. Thanks for explaining it and sorry for making it so painful.
Acked-by: Simon Glass sjg@chromium.org