Re: kern/47563: test cases do not timeout and hang

[julio%meroh.net@localhost]

The following reply was made to PR kern/47563; it has been noted by GNATS.

From: julio%meroh.net@localhost
To: Martin Husemann <martin%duskware.de@localhost>
Cc: gnats-bugs%gnats.netbsd.org@localhost
Subject: Re: kern/47563: test cases do not timeout and hang
Date: Sat, 16 Feb 2013 16:36:34 -0500

 On Feb 16, 2013, at 16:32 , Martin Husemann <martin%duskware.de@localhost> 
wrote:
 
 > On Sat, Feb 16, 2013 at 04:29:12PM -0500, julio%meroh.net@localhost wrote:
 >=20
 >> In all honesty, the code in atf-run that manages timeouts and muxes
 >> the output of the tests is quite retarded.  I won't be surprised if
 >> there are bugs lingering around... but it's hard to explain why such =
 a
 >> bug would only trigger in sparc64 if that's the case.
 >=20
 > No, all is fine, timeout triggers but it does not manage to shut down =
 the
 > test case.
 
 Is the test case sent the term signal?  And the kill signal?  Does it =
 become a zombie?
 
 Is rump_allserver in the same process group as the test case?  Does it =
 receive the signal and die too?
 
 Lastly, maybe you can try changing timeout_callback() in timer.cpp so =
 that it simply reads:
 
 void
 impl::child_timer::timeout_callback(void)
 {
     m_terminate =3D true;
     ::kill(-m_pid, SIGKILL);
 }
 
 That term/kill duality is... unnecessary.=