John Belmonte

> on the other hand it also points out some stuff that might be there intentionally to prevent a typo such as: I'm assuming this new check will be opt-in,...

I tried the patch with my use case and it worked well-- looking forward to the feature. It's been a while since the last rebase, would you run again?

I'm not sure how to go about it, as taxing a tracing JIT would require a lot of code (I think PyPy gives up after a 6000 opcode trace). Some...

It may be useful to write the benchmark against the anyio API, so that it could be run against both asyncio and trio.

I'm thinking about what metrics of my app would most represent the workload. Then create a synthetic benchmark with similar metrics. assuming a single threaded program, not hitting CPU limits,...

here's how my app looks by those metrics: ``` max active tasks: 580 context switch rate: 4,267 / s unique context switch rate (median, 1s window): 84 / s step...

What's missing from the metrics, and especially relevant to JIT implementations, is some notion of the number of opcodes run per step. It should probably exclude repeated loop iterations-- so...

Deleting Library files and `filesharingDirectories` did not help in my case. For me it seems to be related to heavy load in the container, as reported in https://github.com/docker/for-mac/issues/7381

> We use PCM_GAS value to set max acceleration for Honda NIDECs, this does not work for 2020 Odyssey. ~Do you mean PCM_ACCEL?~ (I assume that merged PR commaai/openpilot#22156 was...

I see, it's PCM_GAS https://github.com/commaai/openpilot/blob/3ccf712ac99ce8ef393d87ddd7241d1d823dd5a6/selfdrive/car/honda/hondacan.py#L122 this comment threw me off, referring to it as "PCM_ACCEL": https://github.com/commaai/openpilot/blob/3ccf712ac99ce8ef393d87ddd7241d1d823dd5a6/selfdrive/car/honda/carcontroller.py#L189-L190