[Deron Johnson]

Jesse Barnes writes:

> If possible, the event layer stuff can be abstracted out a bit so that
> the resulting code is at least somewhat maintainable, maybe you have
> ideas for some cleanups that should be done to make that >
> easier/doable?

Probably the best way to do this is to take the functions that I have
modified and make them call through a function pointer. This would
allow us to install the normal version vs. the LG version at runtime
(enabling and disabling LG behavior at run time is an important
requirement.

I was hesitant to make structural changes of this sort, which was
why I followed the XINPUT approach of using ifdefs. But if there
is a consensus on this issue then I can switch over to using
function pointers instead. On the plus side, it will allow us
to locate all of the LG-specific event code together in one
place. On the minus side, it will introduce some additional
indirections into the code, but this code isn't not extremely
performance sensitive anyway. But it will represent the first
major structural change to the event code in almost two decades.

What is the Xorg architectural process for determining the
resolution of this issue. Do you folks call for a vote on it?
And who is included in the vote? And do I have a vote?

I can change the code in whatever way you would like, just
let me know definitively whether you require it to be done because
it is a fair amount of work.