proposal: os/exec: allow CommandContext termination signal to be specified

[caspervonb]

Currently CommandContext will send a SIGKILL which for the common case means there will likely be a dirty exit. It would be nice if we had the ability to either specify or override the kill behaviour.

This has previously been brought up in #22757 and #21135, where it was deemed as not having enough demand and would be revisited upon later.

[bradfitz]

Got a particular proposal? New Kill func(*os.Process) func field on exec.Cmd?

[rsc]

it was deemed as not having enough demand and would be revisited upon later.

We said in #22757 we'd revisit this if more evidence of demand accumulated. Can you give any information about why the demand might have gone up, or new examples of why this is important?

[gopherbot]

Timed out in state WaitingForInfo. Closing.

(I am just a bot, though. Please speak up if this is a mistake or you have the requested information.)

[tooolbox]

Can you give any information about why the demand might have gone up, or new examples of why this is important?

Example: If the command you're executing is go run, using SIGKILL orphans the child process.

Got a particular proposal? New Kill func(*os.Process) func field on exec.Cmd?

Sounds great!

[bcmills]

Got a particular proposal?

Can you give any information about why the demand might have gone up, or new examples of why this is important?

I've filed #50436 based on the last couple years of experience with this within the Go project. It includes links to several specific code examples within the project and links to some associated Go testing bugs.