[nptl] Re: Patch for SCHED_RR test with LinuxThreads

["Loic Domaigne" <loic-dev@gmx.net>]

Hello! 

> > The attached patch will help prevent the computer hard-reboot.
> 
> Note that the sample sometimes hangs even before the 
> "Starting 200 threads SCHED_RR with prio 5" message. In this case, this
> patch is not enough... 

Ok, I see. My program has 3 'major' issues: 

* We must ensure that the [i]alarm_thread[/i] has the highest prio so
  that it preempts all other threads ([i]testdrv_thread[/i] included)
  when the alarm is raised. 
  [ 
    Though I believe, the current design should be ok on machine 
    with NCPU >= 2. 
  ]

* We must ensure that the timer is started before [i]testdrv_thread[/i] 
  starts to spawn the SCHED_RR threads.

* The process might block anywhere, for instance in pthread_create(). 
  That is, the thread info [b]MUST BE[/b] printed in the
  [i]alarm_thread[/i] handler, not in [i]testdrv_thread[/i] !!! 


I have designed a "new version" of the [i]test_schedrr.c[/i]. 
The three points above have been counted into account AND somewhat
more useful info are printed out when the test failed. 

I have new results (which shall not necessarily enlight us, but that's 
another story)... I guess, we have to re-run all the tests. Sorry about
that... 
[
though I don't know if I am really sorry: nobody else found those issues
:-D
]

I shall publish the new version by Monday. I want to give me a few 
days in order to review the program. 


Cheers,
Loic. 

-- 
--
// Sender address goes to /dev/null (!!) 
// Use my 32/64 bits, ANSI C89, compliant email-address instead:   

unsigned y[]=
{0,34432,26811,16721,41866,63119,61007,48155,26147,10986};
void x(z){putchar(z);}; unsigned t; 
main(i){if(i<10){t=(y[i]*47560)%65521;x(t>>8);x(t&255);main(++i);}}

GMX ProMail mit bestem Virenschutz http://www.gmx.net/de/go/mail
+++ Empfehlung der Redaktion +++ Internet Professionell 10/04 +++

Liste de diffusion nptl
Pour se désinscrire : mailto:nptl_request@bullopensource.org?subject=unsubscribe