Closed douglas-raillard-arm closed 1 year ago
So we should probably:
Target.kick_off()
as it really is just a Target.background()
that leaks which is definitively an anti-patternEDIT: the 2) is implemented there but needs an adjustment before being ready: https://github.com/ARM-software/devlib/pull/626
I can confirm that with the updated PR the instrument finally works as expected so simply getting rid of kick_off
this way is probably a good idea.
Following @mrkajetanp tests, I updated the PR to use
Target.background()
instead of relying onnohup
and&
. The&
unfortunately only makes the command a background job in the shell, butadb shell
blocks until all background jobs are finished before exiting, making it useless.Instead, using
Target.background()
provides the expected behavior. The main change is that background commands are canceled on target disconnect for some connection classes (but not all it seems, we might want to align behaviors there)