Hi,
Thanks for nice miner.
I had been using "-watchdog=false" option which gave me a lot of flexibility. With that option, bminer terminates when it encounters some errors. However it seems there is no defined "exit status" code which can be utilized in parent process. If that was already implemented, please let us know.
I don't know CUDA programming at all but there would be many possible error cases. Not all errors are handled as "last resort recovery procedure" defined in parent process. For instance, transient network related errors wouldn't be fatal CUDA errors so parent process can wait for a while or switch to different pools. On the other hand, some errors could be recovered by system reboot only.
Possible exit status code I have in my mind are:
0 - Ok, terminated without error
1 - Command line argument error(i.e. unrecognized arguments)
2 - Terminated by a signal (i.e. SIGTERM and etc)
3 - Network error(i.e. connection establishment failure after retries)
4 - CUDA error
What's your opinion?