On 5/29/20 1:09 PM, Chet Ramey wrote: > On 5/28/20 6:41 PM, Rob Landley wrote: >>> If you're worried about leaving stray background jobs around, end your >>> test script with `wait'. >> >> Which will hang if the background job is hung. > > It's pretty easy to send yourself a signal from a subshell after a period > of time as a cheap alarm.
I can do various forms of cleanup, yes. But individual whack-a-mole will always have something else it misses (you can sing "a signal won't always kill zombies" to the tune of "my bonnie lies over the ocean"), so I want more systematic cleanup. I was looking at containers, but now I'm running the test in a temporary qemu system that not only does cleanup but lets me run test as root in a known environment where I know what processes are running and rivers are loaded and filesystems are mounted and everything. (Hard to test things like "mount" and "modpobe" otherwise.) >> >>> But you can always save a background job's pid >>> by saving $!, so you can always have a handle to the background jobs you >>> create. >> >> In theory jobs -p should do it, in practice I've hit background-from-subshell >> cases where it wasn't in the list... > > Sure, you can defeat anything. I've done stuff like that myself to avoid > processes being killed by the shell at exit. When running over a thousand tests that do weird stuff, things leak. (Then add tests that need to run as root... :) >>> Bash-5.1 will change the way it does $SECONDS to use gettimeofday(), btw. >> >> Meanwhile, I'm fielding: >> >> https://github.com/landley/toybox/pull/210 > > Ironically, gettimeofday() is more portable. I know. I didn't say I accepted the pull request. I just found it, as you said, ironic. :) > Chet Rob _______________________________________________ Toybox mailing list Toybox@lists.landley.net http://lists.landley.net/listinfo.cgi/toybox-landley.net
