db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ole Solberg <Ole.Solb...@Sun.COM>
Subject Re: store/OnlineCompressTest.java fails on Linux-2.4.20-31.9 i686-i686 (RedHat 9)
Date Mon, 06 Jun 2005 08:57:08 GMT
Mike Matrigali wrote:
> I believe svn 180008 should correct this problem, but I was never able
> to exactly reproduce your error - I didn't make it to a linux machine.
> I believe the problem was a timing related error where the system
> was sometimes incorrectly reading past end of file.  On my machine
> I would see an EOF type error every 5-10 times I ran the test.  After
> the change I ran it 100 times with no error.
> 

FYI, in my tests on revisions 179404 and later(including 180008) the 
problem has not shown up.

> By the way just for reference how fast are processors on your machine,
> and how many processors if is multi-cpu?  It seems the error I fixed
> would show up more easily on fast, multi-cpu machines.

The machine where the problem occurred is a
1 X [Intel(R) Pentium(R) 4 CPU 2.66GHz, 512 KB cache]. 1030872 kB Total 
Memory (Linux).

It was *not* seen on e.g.
2 X [Intel(R) Pentium(R) III CPU family 1400MHz, 512 KB cache]. 2055452 
kB Total Memory (Linux), or
2 X [i86pc i386 (AMD Opteron(tm) Processor 244): 1792 MHz, unknown 
cache].  2048 Megabytes Total Memory (SunOS 5.10).


> 
> Mike Matrigali wrote:
> 
>> I will look into this, see if I can reproduce locally.  Note that
>> lock timeouts in the derby.log are expected part of a successful run -
>> but the other 2 are of course wrong.

[snip]

>>
>> /mikem
>>
>> Ole Solberg wrote:
>>
>>> I have observed that, on 1-one of the platforms we're running 
>>> regression tests on, the store/OnlineCompressTest.java fails.
>>> See e.g. 
>>> http://www.multinet.no/~solberg/public/Apache/Derby/index.html/storemore_history.html

>>>
>>> (storeall and derbyall ditto).
>>>
>>> The platform showing the problem is Linux-2.4.20-31.9 i686-i686 
>>> (RedHat 9).
>>>
>>> There seems to be three different behaviours: 
>>> DATA_UNKNOWN_PAGE_FORMAT, LOCK_TIMEOUT, test never terminates.

[snip]

>>>
>>
>>
>>
> 


-- 
Ole Solberg, Database Technology Group,
Sun Microsystems, Trondheim, Norway

Mime
View raw message