Clobbering of slots in MT protocol handling about koreader-base HOT 12 OPEN

It's a design simplification, as this should be an exceedingly rare occurrence in practice.

(Also, we also clobber slots on contact losses on some Kobo protocols later, anyway).

from koreader-base.

Fair enough. I am working with the remarkable and its happening regularly.

I will do by best to handle it further downstream.

Want me to close this?

from koreader-base.

Want me to close this?

Not necessarily, base's issue tracker isn't crowded, so leaving this open isn't really problematic.

As far as I'm concerned, this would be fairly low on the priority list, even if I actually had time to work on this (spoiler: I don't have any time ;p).

from koreader-base.

Could possibly be as simple as passing along the fd number from the C module to Lua-land and simply using fd + ABS_MT_SLOT in GestureManager, as we don't actually care about the slots value per se (in fact, some Kobo devices don't even emit a slot 0).

from koreader-base.

Yeah I did think about something along those lines, but I am not entirely sure what goes on in KOReader wrt slots and ids once they go into gesture detector.

I am just getting some funny double free exceptions from somewhere and it seems to be related to touch and tool events mingling.

from koreader-base.

Might need to be slightly more complex, because of the initial state in https://github.com/koreader/koreader/blob/19a607b5484299623cc57563f7a4baa3f49faff5/frontend/device/input.lua#L220-L226 (among possibly other things).

If we had a guarantee that the first event device we open is the touch screen, we might be able to use the inputfds index instead, but we don't.

from koreader-base.

It would probably not be too gnarly to deal with, I just have zero time to devote to this (or anything else, really :/) for the foreseeable future.

from koreader-base.

ok cool.

I will leave it be.

from koreader-base.

There might be a similar issue with cross-device ABS_MT_TRACKING_ID values, too (although a collision there is possibly much rarer).

from koreader-base.

From my little bit of watching the event streams on this one device then yes, it looks rarer. One device always uses the same ABS_MT_TRACKING_ID 0, and the other just continuously increments it.

from koreader-base.

As far as passing along the fd info, it should only take some minor contortions to stuff it in the table ;).

from koreader-base.

that is probably the simplest unobtrusive fix

from koreader-base.