input_capture portal Signals are not working

[feschber]

The receive_activate() / receive_deactivate() Streams on the input_capture portal never yield any Result.

I did some digging and could narrow it down a bit:

receive_activated calls signal("Activated") which yields a Stream that filter-maps on the result of msg.body::<Activated>().ok()

So I tried to replicated the body() call manually and it always throws the following error:

called `Result::unwrap()` on an `Err` value: Variant(Message("invalid type: character `o`, expected `v`, `a` or `(`"))

Which then of course results in the signal getting filtered by the filter_map call.

This error in body is originated from zvariant::from_slice_fds_for_dynamic_signature -> deserializer_for_signature -> ... -> deserialize_seq where it ultimately fails in zbus/zvariant/src/dbus/de.rs:347 because it does not expect an o.

[bilelmoussaoui]

I think the signature is wrong; so I would replicate the code outside of ashpd, use msg.body::<zvariant::Value>() an then call dbg!(value.value_signature()) and share the output?

[feschber]

Thank you so much! I wanted to get back to this this week but I guess I don't have to!

When I checked the flatpak docs, I totally missed the missing () ...

[bilelmoussaoui]

No worries! the error message could have been better from zbus... Anyways, i will be doing a release shortly that includes this

[feschber]

Im now running into this issue:

thread 'main' panicked at /home/ferdinand/git/lan-mouse/src/backend/producer/libei.rs:122:30:
failed to deserialize body: Variant(PaddingNot0(1))

Do you have an example where you got this to work correctly?

[bilelmoussaoui]

That looks like a completely different issue. You would have to debug it yourself as I am not familiar with libei and don't have the time to dig into it now

[feschber]

sorry I was a bit unclear but it's still kind of the same problem and I still can not receive the Activated signal because of it.
The signature now seems to get built correctly but parsing fails with this error message.
I will look into it and see what I can find.

[bilelmoussaoui]

If you have code I can easily run to test it, I would be happy to do some debugging.

[bilelmoussaoui]

Maybe @whot would have an idea?

[whot]

The signature now seems to get built correctly but parsing fails with this error message.

which error message? The Variant(PaddingNot0)? It's a bit hard to guess without seeing the code you use to parse (this source doesn't have a line 120).

Probably worth checking with dbus-monitor to see what's being sent and comparing this to your parsing code.

[feschber]

one second

[feschber]

https://github.com/feschber/ashpd-mre

this is basically what I've tried so far.

[feschber]

In lan-mouse the libei-producer branch contains my attempts. (but with more irrelevant stuff)

[feschber]

[2024-03-06T23:26:25Z INFO  ashpd_mre] creating input capture session
[2024-03-06T23:26:27Z INFO  ashpd_mre] selecting zones
[2024-03-06T23:26:27Z INFO  ashpd_mre] Zones { zones: [Region(1440, 2560, 2560, 0), Region(2560, 1440, 0, 815)], zone_set: 0 }
[2024-03-06T23:26:27Z INFO  ashpd_mre] selecting barriers: [Barrier { barrier_id: 0, position: (2560, 0, 2560, 2559) }, Barrier { barrier_id: 1, position: (0, 815, 0, 2254) }]
[2024-03-06T23:26:27Z INFO  ashpd_mre] connect_to_eis
[2024-03-06T23:26:27Z INFO  ashpd_mre] enabling session
[2024-03-06T23:26:27Z INFO  ashpd_mre] receiving activation token ...
[2024-03-06T23:26:27Z INFO  ashpd_mre] Msg { type: Signal, sender: UniqueName(":1.58"), path: ObjectPath("/org/freedesktop/portal/desktop"), iface: InterfaceName("org.freedesktop.portal.InputCapture"), member: MemberName("Activated"), body: Signature("oa{sv}"), fds: [] }
thread 'main' panicked at src/main.rs:126:26:
failed to deserialize body: Variant(PaddingNot0(1))
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

An activated signal is clearly arriving but the deserialization fails here (just updated the example to fix the logging)

[whot]

I have no idea tbh, this looks like it's inside the dictionary deserialization but I don't know how this one works. From a quick look at input_capture.rs that bit looks correct, I don't know if you can somehow hook extra debug messages into the deserialization code to figure out what it's trying to do.

[feschber]

found the culprit!