Re: aha

Re: aha

[Paul Haahr]

[ Byron's suggestion to save the status before and restore it after
doing signal processing, which i agree with 100%.]

> This make it impossible (I think) to program something as "do this as long
> as it works but not longer than 10 seconds".

not at all.  instead of writing
        fn false { return 1 }
        fn sigint { false }
        while( ~ $status(0) 0 ) { ... }
just do
	interrupted = false
	fn sigint { interrupted = true }
	while (!~ $interrupted true) { ... }

Re: aha

[malte]

		interrupted = false
		fn sigint { interrupted = true }
		while (!~ $interrupted true) { ... }

To be precise:

		while (!~ $interrupted true) {{ ... } || break }

Malte

Re: aha

[malte]

	
	Wait! $status already gets set asynchronously with signal handlers
	every time they run a command.

Works great! Instead of

	fn sigint { return 1 }
	while( ~ $status(0) 0 ) { ... }

I use

	fn false { return 1 }
	fn sigint { false }
	while( ~ $status(0) 0 ) { ... }

									So perhaps the signal handler code
	should stash away the old $status and restore it after the signal
	handler is done running.

This make it impossible (I think) to program something as "do this as long
as it works but not longer than 10 seconds".

Malte

Re: aha

[Byron Rakitzis]

Wait! $status already gets set asynchronously with signal handlers
every time they run a command. So perhaps the signal handler code
should stash away the old $status and restore it after the signal
handler is done running.

(I didn't mean to open a can of worms, I think setting $status
asynchronously is a really stupid idea, and I am merely suggesting that
this aspect of rc be cleaned up for 1.5.)

Re: aha

[Tom Culliton x2278]

br>	I am inclined to make the two the same, if only because
br>	it simplifies the interface. It also means that you
br>	will be able to set $status from within a signal handler
br>	by using "return", for whatever that's worth.

EEK! Asynchronously? Like between "if ( ! command )" and "echo command
failed with status $status"?  Sounds iffy to me.

OBNiggle - The man page says the following about wait:  "If no pid is
specified, rc waits for >>ANY<< child process to exit."  Which would be
OK because you can always say "while (! ~ $#apids 0) wait", but it is
not the way things really work.  The man page should say: "If no pid is
specified, rc waits for all child processes to exit."

aha

[Byron Rakitzis]

I understand now the complaint about "signal handlers
not being functions"

The problem is that the rc signal handler does not invoke
signal-functions the same way that functions are normally
invoked. So far the only difference I noticed is that
"return" is not handled.

I am inclined to make the two the same, if only because
it simplifies the interface. It also means that you
will be able to set $status from within a signal handler
by using "return", for whatever that's worth.