#393 Short button presses appear only after another press (regression from 0.10.1)

[Andrew Crawley]

I use a USB-UIRT with lirc via the usb_uirt_raw driver to control my HTPC. I recently installed a distro upgrade that updated LIRC 0.10.1 to 0.10.2, and I'm now seeing a behavior where a short button press (one not long enough to trigger repeat) doesn't appear to applications until after the next button press (short or long, doesn't matter).

I spent the afternoon bisecting my way through the master branch and came up with the following commits of interest:

• [f6ea25] irlog: enhance logging.This is the most recent commit that works properly for me
• [beeb4a] lib, lircd: Remove --release support. This is the very next commit, which introduces a bug where button presses cause lircd to get stuck in a loop consuming 100% CPU
• [8f0ba5] lircd cpu usage may jump to 100% This commit fixes the 100% CPU issue, but I'm having the issue with short button presses in all commits after this (including the current tip of master).

I've attached a log file from the most recent commit. In it, I press the sequence "right down right down" on my remote, with each press being short enough to not repeat and with a few seconds between presses. Here are some useful timestamps:

• First button press (right) begins at 19:31:08.883957 and is immediately sent to the client at 19:31:09.047005
• Second button press (down) begins at 19:31:11.594703, ends at 19:31:11.658258, and is NOT immediately sent to the client
• Third button press (right) begins at 19:31:14.996829. At this point, the previous button press is unblocked and sent to the client almost immediately, with the event for the third press following immediately when the third press ends at 19:31:15.160121
• Fourth button press (down) begins at 19:31:18.102054, ends at 19:31:18.165476, and is NOT immediately send to the client
• I Ctrl-C lircd at 19:31:22.238162, and the final event is sent.

Please let me know if there's any additional information it would be helpful for me to collect or troubleshooting I can perform. Thank you!

[Andrew Crawley]

In case it's useful, here's a log of me pressing the same buttons with commit f6ea25, which works correctly for me.

[Andrew Crawley]

OK, I got nerd-sniped in the process of collecting the data I needed to file a good bug and ended up sending a merge request with a fix for this.

[Sean Young]

I've got an different fix see https://sourceforge.net/u/seanyoung/lirc/ci/ticket-393/
can you let me know if that works

[Andrew Crawley]

Yep, your fix takes care of my issue. I wasn't sure that the remaining code was unused, so I left it alone. Thanks for the quick response!