Let's roll our own subproc spawner ("the way the experts would")

[goodboy]

After all the headaches from trying to integrate trio-run-in-process:

• trip doesn't do control-C right... #115, TRIP - one last shutdown hang #96 as local standouts for which I created Ctrl-C / SIGINT fails to cancel parent on first try ethereum/trio-run-in-process#4
• the mess of 3 repos that aren't doing anything novel and overlap each other as per discussion in Is anyone maintaining this project? ethereum/trio-run-in-process#6
• the over complexity of the design and spawning logic (as discussed by me here and confirmed by nathaniel here)
• the divergence of goals of the author from our own

I think it's time to move on to more general pastures 😉

There was some good refactoring and re-design that came out of this in #95 including the discovery that stdstreams if semaphored correctly can allow for pdb++ to somewhat work out of the box (towards #113).

I would like to discuss with anyio and trio core devs about design and goals to eventually factor this functionality into the core async lib as necessary.

[ryanhiebert]

Thanks for this. Yeah, I think the approach you're mentioning makes sense. And, when it's done right, I think it absolutely makes sense for it to be in the core libraries.

[goodboy]

Relevant links to get this rolling:

• anyio subproc support + todo
• discussion about worker pools in trio
• discussion about spawning processes from nurseries
• the tagged subprocess issue list from trio