commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jörg Schaible <>
Subject RE: [nightly build] io id failed.
Date Mon, 21 Aug 2006 07:06:33 GMT
Henri Yandell wrote on Friday, August 18, 2006 5:48 PM:

> On 8/17/06, Jörg Schaible <> wrote:
>> Hi Phil,
>> Phil Steitz wrote on Friday, August 18, 2006 3:28 AM:
>>> Hi Jorg,
>>> Sorry for the latency.  I have been out of pocket this week.  In
>>> answer to your question, I don't know, I just work here ;-)
>> :)
>>> Seriously, I think is a vmware box and the
>>> nightlies are running in a Ubuntu Linux image.  Here is what I get
>>> from uname: Linux vmbuild 2.6.15-23-server #1 SMP Tue May 23
>>> 15:10:35 UTC 2006 i686 GNU/Linux
>> Yeah, I think so, too. The funny part about it, that
> previously the test failed that assumed, that it could create
> a quite good number of unique ids within a time period and I
> really had a hard time to find a good solution to compensate
> backward time shifts from the OS. Now the opposite hapens!
> This time another test fails, because it simply assumes that
> it cannot generate two ids within the same time slice
> (because the generator is configured in this way) ... and is
> fooled by the OS shifting time forward ;-)
>> I'll modify the test to force the exception ASAP ...
> The lesson appears to be not to get involved with timing code :)

Hehehe ... 

> IO's errors are due to timestamping of files on the file system.
> Lang's error is in its time package.

At least it was a useful lesson for me and I will think twice writing and using code in production
code depending on timings. Especially if there is a chance, that the app will run in a VM,
it is higly unreliable.

- Jörg

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message