Wolfgang Denk wrote:
Dear Tom,
In message 4AF339E1.9060809@windriver.com you wrote:
The arguments for using weak are getting weak :P
:-P
Using weak is less relevant with the #ifdef's
But it's the wrong direction your heading. We should get rid of #ifdef's, not add new ones.
With #ifdef's, you have different versions of the code, which increases the multitude of configurations that actually need to be tested. With weak, you have one version of the code only.
To use the status led api, you have to define CONFIG_STATUS_LED anyway. I did not think this added to the configuration space.
The benefit now is that boards that use the led functions do not have to define all of them.
That's just an indication of a broken implementation.
Normally you would provide the weak functions in a central place, where any board configuration which wants can overwrite them, or not.
I am open to ideas on how to kill weak off completely.
You got it wrong.
We want to kill off the #ifdef's.
My vector is obviously pointing in the wrong direction..
Has a general led driver layer ever been proposed ? Something to convert status led for a mixture of #defines and weak symbols to something that had a register function and some file_ops ?
We use status LEDs on many boards, without real issues. It's only AT91 which suffers from this mess.
I withdraw this patch. I will rethink this and come up with something better.
Tom
Best regards,
Wolfgang Denk