I have these logs as well.

Deleted: messages 480 KB

messages 480 KB View Download

Gio, is this something you would expect from BlueZ?

+mcchou, josephsih for bluez expertise

I don't know why BlueZ could be causing something like this.

From the chrome log attached in C#0, the notifications were received by chrome from bluez regularly at about 60 ms interval throughout the testing period (about 7 minutes). Hence, bluez was doing its job faithfully here.

The problem seems to be on some layer above bluez. FWIW, In the chrome log, we saw something this message appearing regularly each time the notification was received:

connection_holder.h(384)] Instance arc::mojom::BluetoothInstance not available.

I am not sure if this is relevant.

Let me know if any additional logs would be helpful.

I work with multiple devices and this is the only one that experiences this issue or the issue with notifications stopping after several minutes.  This is also the only device that does notifications on a regular interval, and not just when something changes.  I'm not sure if that is relevant, but it is the only difference between this device and the other devices I work with that I can think of.

Fyi, it is possible to bypass D-Bus entirely if the notification needs to have really low latency:

https://git.kernel.org/pub/scm/bluetooth/bluez.git/tree/doc/gatt-api.txt#n126

Hi Luiz, thanks for the pointer which is useful for low-latency notification!

I tried this again today and the behavior seems slightly different than it was before.

Now, as soon as I connect the notifications are delayed by about 1-2 seconds.  After a minute or two the notifications start coming in bursts.

This device still seems to work reliably on chrome on my mac.

FWIW, I'm having these notification issues with two devices.  A Parrot FlyPad and a MicroBit. If anyone has one of these devices I can provide some code to reproduce if that is helpful!

The FlyPad is the original device that has delayed notifications.

The microbit behavior is slightly different.  The notifications don't seem to be delayed on the characteristics that notify very frequently.  But, when those characteristics are notifying, notifications on other characteristics don't consistently come through.

I'm not sure if it makes sense to open a new issue though, considering the problem seems to revolve around characteristics that notify very rapidly.

We do have micro:bit devices available so repro code would be very much appreciated.