On 1 Apr 2003, Slaven Rezic wrote: > Ilya Zakharevich <ilya@Math.Berkeley.EDU> writes: > > > On Tue, Apr 01, 2003 at 12:16:32AM +0200, Slaven Rezic wrote: > > > > > Please look at Message-Id: > > > > > <200303281526.h2SFQCx8031405@vran.herceg.de> for an alternative patch > > > > > for the same problem. Does this work for you? > > > > > > The test succeeds, but there is a long (15 sec?) delay before ok 5. > > > > > 11 seconds. The timeout is defined in line 31 in 190_alarm.t. > > > > But *should* it block on my system? It did not before your patch. > > > > Before the patch, the croak("tcp connect error - $!") caused the > eval{} to be exited prematurely. My intention was to have it spin for 11 seconds and make sure alarm() kicked it out of the infinite loop. This would ensure Net::Ping didn't dink with the ALRM timer in any way. Thus alarm() works correctly if it "blocks" for 11 seconds. Now if it hangs for much more than that, like 15 seconds, than something is wrong. -- RobThread Previous