search

qmk / qmk_firmware

Open-source keyboard firmware for Atmel AVR and Arm USB families
https://qmk.fm
GNU General Public License v2.0
18k stars 38.69k forks source link

[Bug] Debounce on ErgoDash regressed somewhere between 0.10.35 and 0.12.18 #12285

Open tcrs opened 3 years ago

tcrs commented 3 years ago

Describe the Bug

I rebased my keymap from 0.10.35 to 0.12.18, then reflashed my board. Now I get a lot of key repeats, especially on the pinkie finger keys, which makes the board almost unusable. If I rebase back onto 0.10.35 and rebuild/flash everything works fine. You can see my keymap here: https://github.com/qmk/qmk_firmware/compare/master...tcrs:ergosnap-isolike - I don't think there is anything "interesting" going on in the keymap.

System Information

Additional Context

I found this issue: https://github.com/qmk/qmk_firmware/issues/6269 Which sounds similar. However the conclusion seemed to be just to fiddle with DEBOUNCE/DEBOUNCE_TYPE until stuff starts working again. I would rather understand why this has changed and how to avoid the same problem happening again in the future. DEBOUNCE is set to 5 in ergodash/rev1/config.h in both 0.10.35 and 0.12.18, and it doesn't look like the default algorithm has changed in common_features.mk (sym_defer_g or sym_g as it used to be named). It doesn't look like there have been any material changes in quantum/debounce in between 0.10.35 and 0.12.18.

Could anyone give me some tips on how to debug what is happening here? It seems like a clear regression which probably doesn't affect only ErgoDash...

makerinmaking commented 3 years ago

I have a DIY redox and I upgraded to 0.12.20 and saw the same issue with debounce not working. Went back to 0.11.33 and it is working fine, with the same config.

tomboland-vocovo commented 3 years ago

I've just built a dactyl manuform, and I was compiling the firmware from the master branch as of about a week ago. I've been having terrible debouncing issues on my slave side. If I build with EE_HANDS, and try both sides, then it's always the slave side which has the double-tapping issues. I'm using TRRS, and I've tried different pins on my pro-micros, and lots of different cables, doubly-triply checked all my soldering etc. I didn't really have a baseline to go on, as it's the first keyboard I've built. In any case, I took @makerinmaking 's comment, and checked out 0.11.33, and voila, no debouncing issues any more!

tomboland-vocovo commented 3 years ago

I'm happy to help where I can. I can try a git bisection perhaps.

tcrs commented 3 years ago

I'm actually just building another keyboard. Once I've finished that I'll have a "spare" to fiddle around with and I'll try and narrow down the range of working vs. non-working versions.

BRagsdale09 commented 3 years ago

I was fighting the same issue and found this issue, thank you for that!. I've tested up to version 11.60, but 12.0 breaks it on my Kyria.

yejianfengblue commented 3 years ago

The last two columns of slave side of my dactyl manuform have debounce/chatter issue. My original problem is posted in reddit.

I follow this github issue to use old version and don't set any configuration. According to my test, version 0.11.69 doesn't have chatter while version 0.12.0 has chatter. I am testing commit by commit.

Below are my test result to version 0.12.50, no idea if it helps, just post it here.

haldiraros commented 3 years ago

I have the same issue on Sol 2. For me it happens with basically every key on the slave side. In my case the chatter appears if I disable RGB. Same as with others 0.11.69 works fine, 0.12.0 doesn't.

Neither enabling console nor changing the debounce type change things for me. Only toggling RGB (rgb matrix) or switching to lighter animations.

I enabled DEBUG_MATRIX_SCAN_RATE to debug a bit.

It seems that for me on newest QMK the chatter on keypresses on slave side happens when the matrix scan frequency goes above 900 when RGB is disabled. With simple animations that take less processing occasional double presses happen when scan frequency is around 850. With more complex ones and lower scan frequency everything works fine.

On QMK 0.11.69 matrix scan frequency reported even higher with RGB disabled and chatter does not happen.

No idea if this correlation of performance and chatter is actually important but I hope it helps

tzarc commented 3 years ago

If you're able, can you try #11930?

haldiraros commented 3 years ago

Tried #11930. Chatter went away. Got 0 repeat keypresses 👍

tomboland-vocovo commented 3 years ago

This is exciting! I'll try it tonight.

tomboland-vocovo commented 3 years ago

This is seeming to work perfectly for me too! I'll keep an eye on this getting merged in, but I'll stick with your fork for the time being. :)

Now, if I could just get my dactyl to work when my machine wakes up, without having to unplug and re-plug it in every single time, life would be perfect!

TonehavenE commented 3 years ago

I had similar issues, tried #11930 and it immediately went away. Nice job 👍