What is Primitive: 18:1 in 2.0.0-alpha.180?
# helpa
addshore
05/19/2025, 9:11 AMGot a stack and would love to know what primitive this is in this version?
Perhaps I should write down how to find this out as it seems to come up a fair bit.
Copy code
Potential deadlock detected:
Process: 3
Program: a4ae897d-01f6-7c9d-fc35-fbe22fd3660d
BCI: 0x3365
Primitive: 18:1
fatal: Potential dead-lockf
floitsch
05/19/2025, 9:12 AMlooking
floitsch
05/19/2025, 9:13 AMEvents::register_monitor_notifierfloitsch
05/19/2025, 9:14 AMWeird... I have never seen a timeout there.
a
addshore
05/19/2025, 9:14 AMinteresting
f
floitsch
05/19/2025, 9:15 AMupdate_resource_monitorLocker locker(mutex_)floitsch
05/19/2025, 9:15 AMHow often does it hapen?
floitsch
05/19/2025, 9:15 AMAny idea what could be causing it?
a
addshore
05/19/2025, 9:16 AMNot sure, but got it reported from a client, I myself havn't seen it.
addshore
05/19/2025, 9:17 AMThe code makes use of Latches and Channels
addshore
05/19/2025, 9:17 AMChannels working as message inboxes / queues for processing
Latches working as things to wait for responses to messages that are potentially being sent
addshore
05/19/2025, 9:17 AMThe channels are rather high throughput
f
floitsch
05/19/2025, 9:18 AMStill weird. But will discuss it with Kasper in our daily later today.
a
addshore
05/19/2025, 9:19 AMI expect its from this code https://github.com/lightbug-io/toit-lightbug/blob/main/src/services/comms/comms.toit
f
floitsch
05/19/2025, 9:19 AM(side-note:
kebab-casea
addshore
05/19/2025, 9:20 AM😛
addshore
05/19/2025, 9:21 AMThe other non public codes does also...
Copy code
channel_ = Channel 20
....
channel-send msg/protocol.Message -> none:
if active_:
channel_.send msg
...
run:
while true:
while msg / protocol.Message? := channel_.receive --blocking=false:
...location-channel / Channel := ?f
floitsch
05/19/2025, 9:29 AMI think the
register-monitor-notifiersleepreceivek
kasperl
05/19/2025, 10:23 AM@addshore Do you think we can get our hands on the snapshot for the program where this fails?
a4ae897d-01f6-7c9d-fc35-fbe22fd3660da
addshore
05/19/2025, 11:42 AM@kasperl I expect so, I'll go and figure that out 🙂
addshore
05/19/2025, 11:43 AMJust the toit files won't be enough right?
f
floitsch
05/19/2025, 12:21 PMThe toit files might be enough, if we compile to the same snapshot, but the actual snapshot would be more convenient.
If you compiled the pod on your machine you can try to find it in
$HOME/local/.state/toit/a4ae897d-01f6-7c9d-fc35-fbe22fd3660d.snapshota
addshore
05/19/2025, 12:21 PMaaah lovely! that might be the pointer I needed!
addshore
05/19/2025, 12:23 PMI'll send that over to the person that was running it to see if they can get the snapshot to me
f
floitsch
05/19/2025, 12:28 PMThe snapshot is also in the pod. So if it's not in the state directory, then you can download the pod and extract it from there.
a
addshore
05/20/2025, 10:48 AMAlso, does it mke sense for me to plop this one on github too to track it there?
f
floitsch
05/20/2025, 10:49 AMGood question. Probably yes. Currently it doesn't look like a 5-minute fix.
a
addshore
05/20/2025, 10:49 AMIll go write it now and summarize what is here so far
addshore
05/20/2025, 10:52 AMk
kasperl
05/21/2025, 11:25 AMUnsurprisingly the snapshot shows that the bytecode that we're stuck in for too long is the
invoke primitive Copy code
13149: register-monitor-notifier_ <sdk>/core/events_.toit:73:1
0/13153 [018] - load local 4
1/13154 [048] - as class RpcRequestQueue_?(125 - 136)
3/13156 [041] - pop 1
4/13157 [089] - invoke primitive {events::register_monitor_notifier}
8/13161 [053] - invoke static throw <sdk>/core/exceptions.toit:39:1
11/13164 [041] - pop 12 Views