One more detail in case anyone builds the code in the link and tries to reproduce. The call to reg.exe will probably succeed with no prompting the first time (the prompt is for value already exists). So it becomes interesting the second time it is run (and all subsequent times).
On Wednesday, July 5, 2017 at 7:06:23 PM UTC-7, Matt Ford wrote: > > I have a customer that launched reg.exe with the /f switch and created a > monster (it looks like reg.exe prompts for user input, reads stdin, sees > the EOF and re-writes the prompt to stdout in a tight loop). > > I've tried several different ways to stop the write and ended up with a > wrapper writer with an interrupt channel and several sleep statements to > allow the goroutines to yield long enough to notice cancel. I'm curious if > the community knows a better way to force these goroutines to yield so I > can stop the process without introducing sleeps? > > https://play.golang.org/p/kDHRQfgR3Y > > The main process is PowerShell and it ends even though Reg continues to > run - I'm not trying to solve that (It may not be ideal, but it is > acceptable as long as we can prevent it from filling the disk with the > stdout file). > > This problem seems to be a combination of this: > https://github.com/golang/go/issues/13155 (runaway child processes don't > allow output to close, hanging command.wait) and the high frequency writes > preventing the timeout signal from being recognized. > -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
