On 2020-06-24 02:46, Lee Jones wrote: > On Tue, 23 Jun 2020, Frank Rowand wrote: > >> On 2020-06-23 14:59, Lee Jones wrote:
< big snip > Thanks for the replies in the above portion. >>>> But yes or no to my solution #2 (with some slight changes to >>>> make it better (more gracious handling of the detected error) as >>>> discussed elsewhere in the email thread)? >>> >>> Please see "[0]" above! >>> >>> AFAICT your solution #2 involves bombing out *all* devices if there is >>> a duplicate compatible with no 'reg' property value. This is a) >>> over-kill and b) not an error, as I mentioned: >> >> As I mentioned above, I set you up to have this misunderstanding by >> a mistake in one of my earlier emails. So now that I have pointed >> out what I meant here by "more gracious handling of the detected >> error", what do you think of my amended solution #2? > > Explained above, but the LT;DR is that it's not correct. I don't agree with you, I think my solution is better. Even if I prefer my solution, I find your solution to be good enough. So I am dropping my specific objection to returning -EAGAIN from mfd_match_of_node_to_dev() when the node has previously been allocated to a device. > >>>>> It also suffers with false positives. >>> >> >> Sorry for the very long response, but it seemed we were operating >> under some different understandings and I hope I have clarified some >> things. > > Likewise. :) >