Say you have :
;
; a define
(define (foo)
(println "Oh Yeah.")
)
;
; and another
(define (foobar)
(dolist (s (sequence 1 10))
(sleep 1000)
)
)
;
; and a timer:
(define (ImaTimer)
(foo)
(timer 'ImaTimer 1)
)
;
; and, some code:
(println "Setting timer...")
(ImaTimer)
(println "Forking...")
(fork (foobar))
;(foobar)
(exit)
The problem is, the timer never fires in the forked process.
But, if you do not fork, changing
The timer is cancelled from a SIGALRM generated by 'sleep' in the forked child process. If you replace the 'sleep' the timer keeps on going.
(define (foo)
(println "Oh Yeah.")
)
;
; and another
(define (foobar)
(dolist (s (sequence 1 10))
(set 'start (time-of-day))
(while (< (time-of-day) (+ start 1000)))
(println "foobar")
)
)
;
; and a timer:
(define (ImaTimer)
(foo)
(timer 'ImaTimer 1)
)
;
; and, some code:
(println "Setting timer...")
(ImaTimer)
(println "Forking...")
(fork (foobar))
~> newlisp timer
Setting timer...
Oh Yeah.
Forking...
newLISP v.10.3.0 on OSX IPv4/6 UTF-8, execute 'newlisp -h' for more info.
> Oh Yeah.
foobar
Oh Yeah.
foobar
Oh Yeah.
foobar
foobar
Oh Yeah.
foobar
Oh Yeah.
foobar
Oh Yeah.
foobar
Oh Yeah.
Would anything else cause this?
I have removed every (sleep) from my code and I still have the timer failing when I fork.
newLISP uses the same fork(), setitimer() and getitimer() functions on all UNIX platforms. Perhaps there are some specifics, how fork() and setitimer() interact on your platform, e.g. some resource contention between the process scheduler and the timer.
Or... a stray (sleep) call in one of the twenty modules.
Not that I'd ever admit it. :)
Quote from: "Lutz"