It's quite interesting to me the way that different "programming cultures" exist around debuggers.
If you grew up doing windows C++ development, looking at things in a debugger is your first step. You only resort to printing values if you can't immediately see what happened in the debugger.
A lot of other envioronment/language cultures are the opposite. Obviously both have their place, but I do feel like more people should use the debugger as the first step instead of the last.
For embedded applications, especially robotics, it tends not to be a great default because it stops the process, which tends to be bad for realtime control loops. That said, a complete hang is the situation where I absolutely would try to get a debugger attached and get a backtrace as one of the first things to try.
This kind of stuff is why devs doing safety critical work often painfully reinvent the wheel. Even if you’ve personally read the code yourself and think you understand it, there’s always some latent defect that arises from someone else’s bad assumptions.
Yeah. Actually, as I read it, I'm not sure if the robot is running WebRTC or not (In my comment I assumed it was)
But yeah it would be much more predictable for everyone if the robot didn't use WebRTC or the fancy logging library, and there was a WebRTC shim on the laptop to get the visuals into a browser.
The longer I think about that 10 ms control loop, the more I hope they aren't running any WebRTC thing on the same hardware cores as the control loop.
Man I miss embedded robotics work. So fun to write a control loop / algorithm and then see it play out in the real world. <robot crashes into wall> Whoops, guess we better review that routine...
The last photo appears to show the view out the author's office in Fort Mason. Didn't know they had offices there, that's quite a nice view of the Bay.
Sometimes I yearn for the Haskell or Idris style of programming where a dependency can do nothing harmful or stupid without me passing in permission.
Then I think about having to pass in thread handles and file handles to logging libraries. I don't know. It would be a cool option. There is probably a hack for `tracing` that would let me manage the logging thread myself.
Software is so complex these days. The funny solution of doing static-allocated C with no threads and no logging isn't gonna work for me. You aren't going to have WebRTC in from-scratch C.
If you grew up doing windows C++ development, looking at things in a debugger is your first step. You only resort to printing values if you can't immediately see what happened in the debugger.
A lot of other envioronment/language cultures are the opposite. Obviously both have their place, but I do feel like more people should use the debugger as the first step instead of the last.
But yeah it would be much more predictable for everyone if the robot didn't use WebRTC or the fancy logging library, and there was a WebRTC shim on the laptop to get the visuals into a browser.
The longer I think about that 10 ms control loop, the more I hope they aren't running any WebRTC thing on the same hardware cores as the control loop.
Sometimes I yearn for the Haskell or Idris style of programming where a dependency can do nothing harmful or stupid without me passing in permission.
Then I think about having to pass in thread handles and file handles to logging libraries. I don't know. It would be a cool option. There is probably a hack for `tracing` that would let me manage the logging thread myself.
Software is so complex these days. The funny solution of doing static-allocated C with no threads and no logging isn't gonna work for me. You aren't going to have WebRTC in from-scratch C.
>Doesn't crash. Doesn't throw. Just ghosts me.
>Same freeze. Same spot. Iteration 1,615. Every single time.
>It's not slow. It's not starved. It's blocked.
>The Reveal
>That's it. That's the fix. 8 hours of debugging. 2 lines changed. Hold the lock for less time. Tale as old as time.
>The Takeaways
>[the way the bulletpoint list is formatted]
pure AI slop. i'm appalled that this obvious garbage is on the frontpage. you even got the title from GPT, didn't you?
You’re welcome to not like the article, and it can even be LLM-assisted, but that doesn’t mean it’s slop