2 messages in net.java.dev.jna.users[jna-users] Non uniform test results ...
FromSent OnAttachments
Roberto CavalcanteJan 16, 2009 2:57 am 
Nikolas LotzJan 16, 2009 4:20 am 
Actions with this message:
Paste this link in email or IM:
Paste this link in email or IM:
Atom feed for this thread
Paste this URL into your reader:
Subject:[jna-users] Non uniform test results over different JDKsActions...
From:Roberto Cavalcante (rmca@gmail.com)
Date:Jan 16, 2009 2:57:27 am
List:net.java.dev.jna.users

Hi,

We did manage to interact with a C DLL. We started to get our responses correct, data was returning OK and so on. But then something changed in one of our machines. It was imperative to remove a software which, of course, there was no doubt it will not damage anything, but it in fact did made a lot of damage. The once functioning solution started to produce non consistent and wrong results. The problem was so strange because it was still working fine in other machines but not that. Then we start to investigate if the problem. First of all, we should be able reproduced it in other machines, but it could not in most of times. It drove us crazy. We were using the latest stable version of JNA, but then we saw you (JNA developement team) had a problem about writing values into memory with inner structures that could be solved in a newer version, then we started to use tag 3.0.9 and we still had the same not constant error results. Four people start to investigate if the problem could no be related to ANSI/UNICODE or alignment stuff in Visual studio, but no conclusion could be trusted because for each new theory there was a new bunch of non constant successes and error results. Then we noticed a pattern related to java version. During tests, some simply crashes with certain java versions. We noticed, that problem was even more catastrofic when the machine have classpath and java home which points to a diferent version of the java available in the windows path. After 3 days the solution (aparently) came, we must be sure about which java version will be available during the execution and it positively does not deliver trustable result with this java version: java version "1.6.0_11" Java(TM) SE Runtime Environment (build 1.6.0_11-b03) Java HotSpot(TM) Client VM (build 11.0-b16, mixed mode, sharing)

We run a 1000 tests (it really was one thousand test, not a fake number) for comparing result with all java virtual machines we have available in our company and some beyond 1.6.0_11 delivered non constant results as well.

We are really interested about problems related to java version you guys may be having.

Please, if we are wrong about our assumption regarding java version, please let us know.

Please comment.

Thanks in advance

Roberto M. Cavalcante