Re: [PATCH v3 1/7] rc: rc-ir-raw: Add scancode encoder callback

["David Härdeman" <david@hardeman.nu>]

On 2015-05-21 13:51, Antti Seppälä wrote:
> On 21 May 2015 at 12:14, David Härdeman <david@hardeman.nu> wrote:
>> I'm talking about ir_raw_encode_scancode() which is entirely broken in 
>> its
>> current state. It will, given more than one enabled protocol, encode a
>> scancode to pulse/space events according to the rules of a randomly 
>> chosen
>> protocol. That random selection will be influenced by things like 
>> *module
>> load order* (independent of the separate fact that passing multiple
>> protocols to it is completely bogus in the first place).
>> 
>> To be clear: the same scancode may be encoded differently depending on 
>> if
>> you've load the nec decoder before or after the rc5 decoder! That kind 
>> of
>> behavior can't go into a release kernel (Mauro...).
>> 
> 
> So... if the ir_raw_handler_list is sorted to eliminate the randomness
> caused by module load ordering you will be happy (or happier)?

No, cause it's a horrible hack. And the caller of ir_raw_handler_list() 
still has no idea of knowing (given more than one protocol) which 
protocol a given scancode will be encoded according to.

> That is something that could be useful even for the ir-decoding
> functionality and might be worth a separate patch.

Useful how?