LineageOS, Current status : 16.0 Test Builds

[npjohnson 33]

7 hours ago, Wheeljack said:

I see a lot of disconnects/reconnects with my WiFi as well. It's even happening while watching a YouTube video. Could this also be aggressive switching between 2.4 and 5GHz?

Edit: I'm also getting a message to please disable airplane mode on my first call attempt after waking my phone.

So, interesting test, can you make your wpa_supplicant_overlay.conf match this? https://github.com/LineageOS/android_device_motorola_nash/blob/lineage-17.1/wifi/wpa_supplicant_overlay.conf

 

See if that helps anything.

[npjohnson 33]

8 hours ago, EskeRahn said:

Despite the more aggressive power management, to my surprise the Test21 is substantially MORE power-hungry than Test20?!? see below from 3C All-in-one toolbox. (It looks similar on the rooted) - note also the x on the WiFi, Mobile data,  the screen dump is taken shortly after unlocking with fingerprint.
Roughly at ratio 7:4 or a factor 175%

And no, there aren't additional stuff running. I checked with Greenify and Android Assistant and only a few user and system apps I explicitly have white-listed are running

This screenshot makes me think you have a persistent VPN (blockada or similar?) can you try without that? Recent QCACLD merge may be the issue - I've heard of issues with persistent VPN's after that.

[EskeRahn 5,460]

1 minute ago, npjohnson said:

This screenshot makes me think you have a persistent VPN (blockada or similar?) can you try without that? Recent QCACLD merge may be the issue - I've heard of issues with persistent VPN's after that.

Well I have NetGuard active all the time, It will be a show stopper for Lineage for me if that no longer works.

[tdm 2,322]

4 minutes ago, npjohnson said:

This screenshot makes me think you have a persistent VPN (blockada or similar?) can you try without that? Recent QCACLD merge may be the issue - I've heard of issues with persistent VPN's after that.

Please take a look at the kernel history on my github.  There was no rebase between test20 and test21 so the kernel history is linear and both tags appear in a simple log.  The qcacld merge was complete for test20.

 

[npjohnson 33]

2 minutes ago, EskeRahn said:

Well I have NetGuard active all the time, It will be a show stopper for Lineage for me if that no longer works.

I understand - I want to get it working - I'm asking purely for testing purposes.

 

If you can't I likely can try some stuff today.

Edited July 15, 2020 by npjohnson

[EskeRahn 5,460]

13 minutes ago, npjohnson said:

I understand - I want to get it working - I'm asking purely for testing purposes.

 

If you can't I likely can try some stuff today.

Well to give you some valid data I would have to flash a phone with Test20, without the VPN, wait a day or so, and then flash Test21, wait a day or so...

So it will clearly be faster to see if someone else without VPN have a battery monitor installed, and see the same difference switching from Test20 to Test21.

[npjohnson 33]

31 minutes ago, EskeRahn said:

Well to give you some valid data I would have to flash a phone with Test20, without the VPN, wait a day or so, and then flash Test21, wait a day or so...

So it will clearly be faster to see if someone else without VPN have a battery monitor installed, and see the same difference switching from Test20 to Test21.

Additionally, can we get your greenify stats?

[EskeRahn 5,460]

4 minutes ago, npjohnson said:

Additionally, can we get your greenify stats?

Sure, how do I pull those?

[tdm 2,322]

2 hours ago, tdm said:

I'm going to go over Ethan's changes today. My first thought is to remove all the power related changes and just take the cleanup changes until we test more thoroughly. I'll post later after I discuss with him.

Actually, scratch that.  I've got other stuff I need to do today and Nolen just received his Pro1 an hour ago.  So I'll let him have a go at this for a while.  He's already got several things fixed, including fixing the display density, enabling the camera button by default, and fixing the fingerprint location image in the setup wizard.  He should be able to track down the WiFi and power issues also.

 

[npjohnson 33]

4 minutes ago, EskeRahn said:

Sure, how do I pull those?

I suppose whatever compilation of data between the app you screenshotted and greenify that provides:

- percentage drain over time

- Suspend time

- Wake-locks

- And anything else you think might be of pertitnence

[EskeRahn 5,460]

Just now, npjohnson said:

I suppose whatever compilation of data between the app you screenshotted and greenify that provides:

- percentage drain over time

- Suspend time

- Wake-locks

- And anything else you think might be of pertitnence

I have no idea how to pull that info. At best I can give you what GSAM can tell...

[EskeRahn 5,460]

....sorry for the random order it ended in.

Note that GSAM has a bug, so any traffic through VPN is logged as WiFi, even if you turn off WiFi completely.

[Wheeljack 212]

2 hours ago, npjohnson said:

So, interesting test, can you make your wpa_supplicant_overlay.conf match this? https://github.com/LineageOS/android_device_motorola_nash/blob/lineage-17.1/wifi/wpa_supplicant_overlay.conf

 

See if that helps anything.

Could you point me to the directory where this file is supposed to be on my phone please? 😅

Edit: okay... found it in /vendor/etc/wifi

Okay... added 6 lines since the first 3 were already there. Rebooted and checked, if the modification was still there. Will keep an eye on WiFi behavior.

Edit 2: So WiFi is now indeed way more stable for me. I have yet to see the WiFi symbol with an X and mobile data activating LTE. There were still a few hiccups with youtube though briefly showing offline/reconnected during playback.

Edited July 15, 2020 by Wheeljack

[tdm 2,322]

Nobody said anything about test20 having any issues, but it also had the qcacld merge (the WiFi driver update in the kernel).  I'm wondering if someone could download and flash the test19 and test20 boot images and see how they work with the test21 build.

 

[EskeRahn 5,460]

56 minutes ago, tdm said:

Nobody said anything about test20 having any issues, but it also had the qcacld merge (the WiFi driver update in the kernel).  I'm wondering if someone could download and flash the test19 and test20 boot images and see how they work with the test21 build.

I saw no substantial differences going from 19 to 20, but did from 20 to 21.

[tdm 2,322]

26 minutes ago, EskeRahn said:

I saw no substantial differences going from 19 to 20, but did from 20 to 21.

Good to know, thank you.  That implies that the qcacld updates are fine and it is Ethan's power stuff that is causing problems.  On the other hand, @npjohnson is seeing the WiFi stabilize when he takes test21 and backs out the qcacld updates.  So the culprit is still not entirely clear.  Unless perhaps both are causing issues...?

 

Edited July 15, 2020 by tdm

[tdm 2,322]

FYI ... for better/quicker battery drain info, look at this file:

 

/sys/devices/soc/800f000.qcom,spmi/spmi-0/spmi0-02/800f000.qcom,spmi:qcom,pmi8998@2:qpnp,fg/power_supply/bms/charge_now_raw

 

I'm not sure what the units are, but I'm charging my battery to 100% now to get an idea.  It's roughly one million at 52% charge.

 

You can check that value, do something to test (like leave the phone to sleep) for about 10 to 30 minutes, then check the value again.  Divide by the seconds elapsed and you have a drain rate.  You could even script it.

 

Edit: the value maxes out at 2499526 on my device. So that's roughly 793 per mAh. Yours may or may not be different.

 

 

Edited July 16, 2020 by tdm

[tdm 2,322]

Ethan believes that he has tracked down the problem (hopefully the only one). I'll build and test and hopefully have a test22 build on the morning (12 hours from now).

[EskeRahn 5,460]

4 hours ago, tdm said:

/sys/devices/soc/800f000.qcom,spmi/spmi-0/spmi0-02/800f000.qcom,spmi:qcom,pmi8998@2:qpnp,fg/power_supply/bms/charge_now_raw

There should be a charge_full_design and charge_full also, that might help the calibration.

Found this googling https://dev.bettermobileapp.com/article/10435402/-+removed+-

 

[marmistrz 93]

As for GitHub issues, we probably should have a single keyboard issues metabug. I've noticed several issues with the current state of the hw kbd support, some of them are scattered across several issues, some not reported yet

• no way to turn off autocapitalization
• fn isn't sticky (but probably should)
• double shift acts as caps lock (but shouldn't, because we have a hw caps lock key)
• custom keymaps are broken under LineageOS: fn & alt are not respected. For instance, if the kcm maps alt+e/fn+e to ę, the sign ´ will be emitted anyway
• alt opens an emoji drawer (it's wrong, the keyboard label is fn+alt)
• currently alt acts as compose, but this probably should be configurable (for instance: alt for diacritic signs, super for compose)

I'm willing to help if you guide me where to start hacking 🙂

Edited July 16, 2020 by marmistrz

[dicer 36]

21 minutes ago, marmistrz said:

 

• shift is sticky (but shouldn't, because we have a hw caps lock key)

For me it is sticky for one character after pressing shift. I think this is perfect for reaching the character you want and it should stay like that. That's also how it has been with LineageOS on the Droid4

[marmistrz 93]

39 minutes ago, dicer said:

For me it is sticky for one character after pressing shift. I think this is perfect for reaching the character you want and it should stay like that. That's also how it has been with LineageOS on the Droid4

Oh, I meant double shift as caps lock. I'll edit my post.

[Wheeljack 212]

Sorry for the big delay between my updates:

after taking the phone off the charger last night, it was not connected to WiFi and mobile again and it took on the same behaviour as before the changes to wpa_supplicant_overlay.conf.

After turning on the phone this morning, WiFi was still spotty, so I tried flashing the the test19.img bootloader as suggested. Phone wouldn't start at all. Went back to test21.img and I'm relieved nothing is broken 😅

Now waiting for test22.

[SteffenWi 139]

Test21: Something weird is definetly going on with the WLAN connectivity. I don't have a SIM card in my Pro1 yet, so I'm only connected via WLAN. I setup the phone on test20, then did some tests (all fine) including having the phone in standby for ~4 hours and waking it up again. All good. Now on test21, after having the phone in standby for 4 hours and waking it up again it was connected to an open WLAN that has a worse signal strength than my own and I have never been connected to that network.

Once I opened WLAN settings I had to disconnect from that open WLAN and connect to my own. Once that was done, I pressed the home button, the settings screen disappeared and the WLAN symbol disappeared as well. Upon opening WLAN settings again, I saw that I was disconnected. After connecting again, I switched over to Firefox and began browsing through news.google.com, clicked an article and right at that moment my WLAN was disconnected again.

 

As for power usage, I used the app "AccuBattery". It splits into several categories (screen on, screen off, deep sleep and discharge speed in %/hour for screen on and screen off).

Power usage with screen off is definetly lower in Test 21 with 0.3%/hour, while Test 20 had 0.5%/hour - this may be due to Wifi shenangians, by which I mean that being connected to a WLAN that has no internet access probably results in less power drawn and the phone may also have been disconnected entirely for some time while it was in deep sleep.

Power usage with screen on is nearly twice as high though for Test21 with 21.8%/hour while Test20 had 12.4%/hour

 

These values were obtained over a 5 hour time window in both cases with similiar uses (if anything I used test20 more because I may have fallen victim to the ol' reddit trap for a few minutes): Youtube in Firefox for ~1:30 minute, browsing the internet (just things like reddit, news.google.com, imgur...) for ~5 minutes, letting the phone sleep for 4 hours and then repeating the previous steps.

Edited July 16, 2020 by SteffenWi

[Hook 3,020]

For what it's worth, I am just not seeing any problems with wifi connectivity and, although I am speaking anecdotally as I don't do much real battery monitoring, I am not noticing any unusual power drain.  I am using Test21. I am in the US with a Qwerty model on the off chance that might make a difference.