Gracefully ending a program from console

[AKASteve]

Is there any way to request a program to exit gracefully from an SSH client?

I've got XBMC running in raspbian and so far the only way I can see to exit it is to manually navigate to the exit button and tell it to end. Does anyone know if I can request for it to exit gracefully from SSH and if so how? I recall reading somewhere that one should exit XBMC before shutting down the PI otherwise you run the risk of corruption and I would rather avoid that.

[elatllat]

man kill

[AKASteve]

Will using the kill command simply end the program or tell the program to go through normal shutdown operations?

[PeterO]

Will using the kill command simply end the program or tell the program to go through normal shutdown operations?

It will depend if the program handles the "SIGTERM" and "SIGINT" signals properly.

Have a look at http://en.wikipedia.org/wiki/Unix_signal fpr some background on unix signals.

PeterO

[DougieLawson]

It will depend if the program handles the "SIGTERM" and "SIGINT" signals properly.

How about using SIGUSR1 or SIGUSR2 and giving those non-terminating signals a special function.

[PeterO]

It will depend if the program handles the "SIGTERM" and "SIGINT" signals properly.

How about using SIGUSR1 or SIGUSR2 and giving those non-terminating signals a special function.

That would be fine if the OP was asking about making their own s/w deal with signals but they are asking about XBMC and I doubt they want to make their own custom version of that If it already correctly handles SIGINT then that is a much easier solution !

PeterO

[Joe Schmoe]

The short summary to this thread is that a well-written piece of software that needs to do some "clean up" before exiting [*] should setup signal handlers for the commonly given signals (and document which ones it handles) so that it can be killed and the right thing(s) will happen.

Whether or not XBMC (or any other software) meets this criterion for quality can only be determined empirically.

[*] On the assumption that if the program just exits (the default action for most signals), that the system would be left in some kind of "bad" state.

[jojopi]

I recall reading somewhere that one should exit XBMC before shutting down the PI otherwise you run the risk of corruption and I would rather avoid that.

Corruption of what?

Anyway, provided you shutdown cleanly (such as "sudo halt") all processes do get a catchable signal and a few seconds to clean up before they are forcibly killed. If this is not sufficient for a given application then it is buggy and killing it by hand may not work any better.

[PeterO]

The short summary to this thread is that a well-written piece of software that needs to do some "clean up" before exiting [*] should setup signal handlers for the commonly given signals (and document which ones it handles) so that it can be killed and the right thing(s) will happen.

+1
PeterO

EDIT: A quick GOOGLE seems to suggest that hadling SIGTERM has been considered and a patch provided for situations when XBMC has "got stuck" thus stops a mediaplayer from doing a shutdown properly.

[PeterO]

I recall reading somewhere that one should exit XBMC before shutting down the PI otherwise you run the risk of corruption and I would rather avoid that.

Corruption of what?

Maybe databases and the like that hold metadata for the media files ? (Just a guess)
PeterO