#162 Redesign --uinput to separate service

[Alec Leamas]

The code for --uinput it's basically an exception in each and every cased it's used. It contains a lot of specific code paths for some functionality which seldom is used.

This should be redesigned so that --uinput option is deprecated and replaced with a new lircd.uinput service. This should connect to lircd as a regular client and forward the data to the kernel uinput device. This design will be much cleaner and will resolve a lot of problems. It will cost an extra process context switch, which all-in-all.

Cleaning-up the overall design will also make a future "lirc 2.0" more feasible.

[Bengt Martensson]

Good idea.

Note that with this change, the main argument for naming remote keys after linux input events will no longer be valid ;-).

[Alec Leamas]

Note that with this change, the main argument for naming remote keys after linux input events will no longer be valid ;-).

I don't think so. First, the main argument is not about uinput, it's about that lircrc files should be independent of the device (i. e., lircd.conf) used. Secondly, the uinput argument is still valid - making this separate service does not really change anything.

[Alec Leamas]

Fixed, part 1, in [ee6d9b]. The remaining work to actually remove the uinput code in lircd must wait at least a release cycle.