This commit re-enables multiple users for all devices. That's possible
as the system image is too large for Android Go nowadays anyway,
therefore no Android Go builds are available anymore.
This commit also increases the maximum number of users from 4 to 5 as
that's what several hardware already overlays set.
This enables links from the settings to the cell broadcast app for
configuring the behavior for received cell broadcasts.
While most hardware overlays already enabled this setting, it's nothing
hardware related, so it makes sense to set here globally for any device.
I'll also open a PR in `vendor_hardware_overlay` to remove the setting
from all overlays there.
This disables audio effects for the Motorola Moto G7 Play (channel) as
enabled audio effects break audio via Bluetooth.
It requires a separate if-statement in `rw-system.sh`, because
`ro.vendor.build.fingerprint` isn't set for the device.
Relates to: https://github.com/phhusson/treble_experimentations/issues/857
* For some reason, the OpenGL implementation on SPRD devices causes
crashes in system_server on glFinish() due to double-free of mutexes.
It is not the same case with the MTK skia compatibility bug, because
this one produces no GL errors, only the final mutex-related crash.
* This crash is only observed in system_server by far. In other
hardware-accelerated applications, after a lot of tinkering, I could
not reproduce the same problem. (Even if they somehow do crash, it
won't bring the entire system down like `system_server`.)
* Since all SPRD devices do not have "high-end gfx", it is reasonable
to set this `ro.config.avoid_gfx_accel` flag. This will disable the
use of hardware gfx in system_server, avoiding the crash. Most SPRD
devices comes with Android Go by default, which has this flag set
(implied by `ro.config.low_ram`) anyway.
* All devices having the `sprd-adf` module is an SPRD device launched
with Android 9 or earlier. ADF is deprecated in Android 10 or higher.
* Somehow, when running GSI 10, Spreadtrum's adf driver do not show up
with the correct name in /dev. I am not sure how this happens, but it
seemed that a few `mknod`s fixed the problem. This was tested on
Xiaomi Qin 2 Pro.
