Re: gettimeofday calls under Redhat ES 4 Update 2

--=-5XNRUWcw4xg0lyGm0/IU
Content-Type: text/plain
Content-Transfer-Encoding: 7bit

On Tue, 2006-02-07 at 13:34 -0500, Dominic Duval wrote:
> Hi,
> 
> Have you tried running the whole thing through lstrace? That might help
> determining which function in glibc causes the call to gettimeofday(). 
> 
> If you have access to a test system, you might also want to downgrade
> glibc to whatever version we used in U1, just to make sure the problem
> comes from glibc (which is what I suspect).
> 
> Let us know if you find the solution.
> 
> Cheers,
> 			-Dominic

Hi,

I have been able to reproduce this with a simple test script (attached).

ltrace[1] demonstraces the change in system call activity. I have
included only the top 5 in each list for brevity:

Under RHES4.1:

     46 SYS_stat
     55 SYS_futex
     58 SYS_lstat
     76 SYS_open
    105 SYS_mmap

Under RHES4.2:

     54 SYS_futex
     58 SYS_lstat
     76 SYS_open
    105 SYS_mmap
   1658 SYS_gettimeofday

Over 1500 calls to gettimeofday, out of thin air.

We've not yet tried a downgrade of glibc on the affected systems, but
that is out next test.

As this is clearly not a problem related to the Blackdown JVM itself,
this shall be my last email to the list on the subject. If any reader is
also affected by this problem and would be interested in updates, by all
means make yourself known to me privately.

Ollie

1: ltrace -S /usr/java/j2sdk1.4.2_02_bd/bin/java BFThreadTest 2>&1|awk
-F\( '{print $1}'|sort|uniq -c | sort -n | tail -5

--=-5XNRUWcw4xg0lyGm0/IU
Content-Disposition: attachment; filename=BFThreadTest.java
Content-Type: text/x-java; name=BFThreadTest.java; charset=utf-8
Content-Transfer-Encoding: base64
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==


--=-5XNRUWcw4xg0lyGm0/IU--


----------------------------------------------------------------------
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]