[00:24] <adlr> I'm not sure if this is a good place to ask: When sending ABS_MT events to userspace, the kernel only sends information that's changed, effectively delta compressing the information
[00:24] <adlr> Does anyone here know where the code is that does this compression? I've been grepping for like an hour
[00:33] <cnd> adlr: where in the kernel?
[00:33] <adlr> yeah
[00:33] <cnd> I would guess somewhere in drivers/input/evdev.c
[00:34] <cnd> why do you want to know though?
[00:34] <adlr> debugging a case where I didn't get  SYN_DROPPED when I should have
[00:35] <cnd> adlr: looks like it's here: http://lxr.linux.no/linux+v3.5.3/drivers/input/input.c#L166
[00:36] <cnd> it returns INPUT_IGNORE_EVENT if *pold == *pval
[00:36] <cnd> it's specific only to absolute events though
[00:36] <cnd> as it should be
[00:36] <adlr> yeah, makes sense. thanks very much!
[00:36] <cnd> np
[19:54] <cnd> bregma: would you be able to review this asap? https://code.launchpad.net/~chasedouglas/geis/fix-geisv1-class-ids/+merge/123354
[19:54] <cnd> dandrader is out
[19:54] <bregma> OK
[19:55] <cnd> hmm… the other option is to say that we've committed to a specific abi
[19:55] <cnd> and we could just change the implementation back to match these hardcoded values
[19:56] <cnd> that way we don't have to rebuild any clients, or force them to build against a specific version of geis
[19:56] <cnd> bregma: what do you think?
[19:57] <bregma> it is an ABI change: existing clients would need to be recompiled for it to work
[19:57] <cnd> yeah
[19:58] <bregma> to avoid an ABI change the new values would have to be mapped to the old values in geis_v1.c
[19:58] <cnd> yeah
[19:58] <cnd> it's a simple change
[19:58] <cnd> I'm leaning towards that instead now
[19:58] <bregma> the proposed change would have subtle breakage with existing clients
[20:21] <cnd> bregma: please review the MP again