>Dan Mick wrote:
>[CC:'ing ksh93-integration-discuss at opensolaris.org]
>> Someone just noticed that sleep hangs if the network is down.
>
>Can you send a stacktrace of the hanging "sleep", please ?
>
>> because, of course, it's doing a nameservice lookup. Of course sleep is.
>> Why
>> wouldn't it?
>
>Uhm... ksh93 has builtin networking functionality for UDP/TCP/SCTP but
>AFAIK this is a completely different codepath which the ksh93 "sleep"
>builtin never uses...
>
>> Um, I'm guessing this is another artifact of being a ksh93 builtin? Does
>> this
>> strike anyone else as a rather large liability of such tricks?
>
>What do you mean ? AFAIK the "sleep" builtin shouldn't do that...
Clearly it does:
truss -t open sleep 1
open("/bin/.paths", O_RDONLY) Err#2 ENOENT
open("/usr/sbin/.paths", O_RDONLY) Err#2 ENOENT
open("/usr/ccs/bin/.paths", O_RDONLY) Err#2 ENOENT
open("/usr/openwin/bin/.paths", O_RDONLY) Err#2 ENOENT
open("/usr/openwin/demo/.paths", O_RDONLY) Err#2 ENOENT
open("/usr/ucb/.paths", O_RDONLY) Err#2 ENOENT
open("/usr/lib/nis/.paths", O_RDONLY) Err#2 ENOENT
open("/opt/etc/.paths", O_RDONLY) Err#2 ENOENT
open("/etc/.paths", O_RDONLY) Err#2 ENOENT
open("/usr/lib/.paths", O_RDONLY) Err#2 ENOENT
open("/var/run/name_service_door", O_RDONLY) = 3
open("/dev/null", O_WRONLY|O_CREAT|O_TRUNC, 0666) = 4
Why?
Casper
