[9fans] _SLEEP.s

[9fans] _SLEEP.s

[lucio]

When attempting to debug tcl 7.6, one of the tests that fails uses
select() and seems to get caught in an infinite loop.

Acid reports:

acid: lstk()
_SLEEP()+0x7 /sys/src/ape/lib/ap/syscall/_SLEEP.s:5
select(timeout=0x5d2d0,rfds=0x0,wfds=0x0,efds=0x0,nfds=0x0)+0x5e3 /sys/src/ape/lib/ap/plan9/_buf.c:280
	t=0xa
	f=0x7fffd3c4
	i=0x8a43a26c
	n=0xe9d6f
	slots=0x0
	err=0x7fffd3bc
Tcl_Sleep(ms=0xa)+0xe8 /pub/src/tcl7.6/plan9/tclPlan9Notfy.c:319
	before=0x8a43a26c
	after=0x8a43a26c
[ ... ]

and what baffles me is that the stated syscall code is not to be found
in the given location.  The nearest I came up with was in the
.../gnu/ape hierarchy.

I checked sources and that also has no entries other than the mkfile
in /sys/src/ape/lib/ap/syscall so perhaps somebody can explain how
that mkfile works, at least in broad terms.

Lastly, can anyone hazard a reasonable guess as to how the sleep
syscall actually gets itself in a loop?  The timeout value is "10", if
acid is not lying.

++L

Re: [9fans] _SLEEP.s

[C H Forsyth]

the mkfile generates the .s files automatically using
a cunning for loop on syscall.h and case on $objtype.
it's hygienic because it tidies up right away.