Use notifications instead of polling for determining battery status about solaar

[pwr]

HID++ 2.0 has support for a battery feature. Not all HID++ 2.0 devices have that feature, however: e.g. the K750 keyboard does not, but has a different feature related to the light sensor that also reports the battery charge.

[pwr]

Right now Solaar tries to enable all notifications, but on the receiver alone. I haven't looked yet at notification flags on the devices themselves.

[Lekensteyn]

Just do not enable all notifications by setting FF FF FF. On the k800, r1_bit1 has a side-effect that some keys (like Mute) are not delivered to the OS (https://git.lekensteyn.nl/ltunify/tree/registers.txt#n81). It should be possible to make Solaar generate events though... but that is getting off-topic in this issue I guess.

[pwr]

That is interesting to know; will have to check if it has the same effect on my K750.

[Lekensteyn]

Please have a look at https://github.com/Lekensteyn/Solaar/commits/sandbox, mainly:
Lekensteyn/Solaar@742d5de - RFC battery notification processing.
Lekensteyn/Solaar@e3a2731 - (simple commit moving battery register parsing into a separate function)
(the other commit is a simple fix that can be cherry picked if you want).

TODO: enable device notification enablement, can you suggest a place to do so? A quick look did not reveal a central place where new devices are (un)tracked.

[pwr]

Okay, I've done some testing with my M705 mouse; it's 0x0D register behaves the same as the 0x07 register, when the battery_notification flag is set. The battery status mesage (10 0x 0D) is only sent once, when the flag is set. No idea if it is sent at any point afterwards (the battery discharges very slow).

[pwr]

In lib/solaar/listener.py, line 141 (in _notifications_handler()), is where the devices are properly set-up after being discovered.

[Lekensteyn]

Heh, I wish we had more HID++ 1.0 devices to check. I see a Performance MX mouse in the repo, do you have that device in your possession? Btw, my M525 has the same "problem", it is still full ("90%") after one year and I cannot get this down (the Illuminated keyboard is easier to empty, set lights to a max brightness and wait)

[pwr]

No Performance MX, just a K750 and M705.
However, I can borrow tomorrow a M515 to check its registers.

[Lekensteyn]

That would be great. So far, the common registers seems to be:

• 00: Notifications (m510, k800, k705; should apply to every machine, in HID++ 1.0 spec)
• 01: different functions on m510, k800, k705
• 07: battery status (m510, k800)
  -- D0: unknown (k800, k705); different value on m510
• F1: firmware/bootloader version (m510, k800, k705)
• F3: unknown function (m510, k800, k705)

Hopefully the differences exist only between types (keyboards, mice, ...) and not models.

[pwr]

Hm, the M515 is a HID++2.0 device, and has the standard BATTERY feature (no battery registers).

[terwarf]

I own a Performance MX mouse - if you tell me what to do, i would love to help in these issue...

[Lekensteyn]

@terwarf Great, please turn off interfering programs (like Solaar) and run the tools/scan-registers.sh tool from git with Python 2 as python in your PATH. Wait a minute for completion and paste the output here.

[Lekensteyn]

@pwr I see you did some work for integrating other types of notifications. How is the battery status supposed to be updated? By writing self[BATTERY_LEVEL] and calling self._changed()?

[pwr]

Yup, the first if of _process_feature_notification() is a pretty good example.

[Lekensteyn]

I was (and am still) confused by this:

self[BATTERY_STATUS] = BATTERY_STATUS[battery_status]

BATTERY_STATUS = 'battery-status' in status.py and _NamedInts in hidpp20.py. Shouldn't it be:

self[BATTERY_STATUS] = _hidpp20.BATTERY_STATUS[battery_status]

[pwr]

Yup, that's a bug.
This is what happents when I don't have a device with the feature, to trip over stuff like this :).

[Lekensteyn]

I will fix it together with refactoring this in a function in preparation for #30.

[terwarf]

Here is the output of scan-registers.sh from my Performance MX mouse:

alex@linux-v34d:~/Downloads/Solaar-master/tools> ./scan-registers.sh 1
<< (   0.017) [10 01 8100 000000] '\x10\x01\x81\x00\x00\x00\x00'
>> (   0.049) [10 01 8100 000000] '\x10\x01\x81\x00x00x00\x00'
<< (   1.050) [10 01 8101 000000] '\x10\x01\x81\x01\x00\x00\x00'
>> (   1.097) [10 01 8101 020000] '\x10\x01\x81\x01\x02\x00\x00'
--
<< (   7.335) [10 01 8107 000000] '\x10\x01\x81\x07\x00\x00\x00' 
>> (   7.382) [10 01 8107 070000] '\x10\x01\x81\x07\x07\x00\x00'
-- 
<< (  86.592) [10 01 8151 000000] '\x10\x01\x81Q\x00\x00\x00'
>> (  86.639) [10 01 8151 111100] '\x10\x01\x81Q\x11\x11\x00'
--
<< ( 108.430) [10 01 8163 000000] '\x10\x01\x81c\x00\x00\x00'
>> ( 108.477) [10 01 8163 890000] '\x10\x01\x81c\x89\x00\x00'
--
<< ( 240.505) [10 01 81D0 000000] '\x10\x01\x81\xd0\x00\x00\x00'
>> ( 240.550) [10 01 81D0 000000] '\x10\x01\x81\xd0\x00\x00\x00'
--
<< ( 245.690) [10 01 81D4 000000] '\x10\x01\x81\xd4\x00\x00\x00'
>> ( 245.737) [10 01 81D4 000012] '\x10\x01\x81\xd4\x00\x00\x12'
--
<< ( 281.016) [10 01 81F1 000000] '\x10\x01\x81\xf1\x00\x00\x00'
>> ( 282.177) [10 01 8F81 F10300] '\x10\x01\x8f\x81\xf1\x03\x00'
-- 
<< ( 284.106) [10 01 81F3 000000] '\x10\x01\x81\xf3\x00\x00\x00'
>> ( 284.153) [10 01 81F3 000000] '\x10\x01\x81\xf3\x00\x00\x00'

[Lekensteyn]

Odd, your notifications (the second line at 0.049) seem to miss some slashes, bug?

Please show the output of the following, it is on the tool-updates branch of my repo:

dev=1
do_req() { tools/hidconsole --hidpp | grep -v ' 8F.. ..0[12]' | grep -B 1 '^>> '; }
reg00=$(echo "10 0$dev 8100 000000" | do_req)
oldflags=$(echo "$reg00" | grep -Po '>>.*? 8100 \K[0-9a-f]{6}(?=\])')
if [ -n "$oldflags" ]; then
        echo "# Old notification flags: $oldflags"
        {
                echo "10 0$dev 8000 ffffff"     # enable all notifications
                echo "10 0$dev 8100 000000"     # read available notifs
                echo "10 0$dev 8000 $oldflags"  # restore notifications
        } | do_req | grep -B 1 '^>>.* 8100 '
fi

It should looks like:

# Old notification flags: 000000
<< (   1.098) [10 02 8100 000000] b'\x10\x02\x81\x00\x00\x00\x00'
>> (   1.347) [10 02 8100 130200] b'\x10\x02\x81\x00\x13\x02\x00'

(if it is empty, your device is incorrect. Try 2 or 3 depending on the connected machines)

[terwarf]

That's probably my fault. I copied the output from console and have reformatted it to get rid of the whitespaces...

The Code above gives me:

# Old notification flags: 000000
<< (   1.113) [10 01 8100 000000] '\x10\x01\x81\x00\x00\x00\x00'
>> (   1.157) [10 01 8100 100000] '\x10\x01\x81\x00\x10\x00\x00'

[Lekensteyn]

Thank you, it looks like only battery updates are available from notifications.

[pwr]

The notification flags are now enabled for each device as it becomes available.