Dean Lee

> How's the can and sendcan performance on the capnp side? Any bottlenecks there? @adeebshihadeh : I don’t think there are bottlenecks in capnp or msgq. I’ll run some tests...

The table below shows the average message sizes calculated using a simple Python script from the demo route. | Message Type | Average Size (bytes) | |-----------------------------------|----------------------| | sentinel |...

> 👀 > > ```python > batman@workstation-shane:~/openpilot/selfdrive/debug$ ./check_can_parser_performance.py > 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 10/10 [00:01 6000 CAN packets, 10 runs > 99.36 mean ms, 99.86 max ms, 98.82 min ms, 0.33 std ms...

> > After this PR, we can simplify car/interface by using only one CanParser. With a single call to update_string for parsing all messages, there's no need to repeatedly feed...

This PR is too complicated. Closed. Some functions have been moved to [PR #1382](https://github.com/commaai/opendbc/pull/1382).

I’m unable to reproduce this issue on my device. @incognitojam : Could this be related to a timing issue on device? The latest code uses `rl.get_frame_time()` to adjust the rotation...

> scrolling doesn't seem smooth, it ignores ~half of your fast scrolling input We don't handle scroll events directly in Cabana. The issue is likely due to Qt's default behavior...

> Maybe a ./cabana.py Python wrapper that finds the URLs then passes into Cabana? So you don't have to re-write it in C++ Modified logreader.py to optionally output log identifiers...

> It would be nice to have a reset button that goes back to the latest datapoints in the stream after you've paused and zoomed. It's difficult to click constantly...

@sshane I didn't find issue when reviewing the code. Running `zmq=1 ./junggle.py --stream` also appears to be normal.Has this problem always been present, or is it intermittent?