db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Knut Anders Hatlen <knut.hat...@oracle.com>
Subject Re: lots of regression test errors on Java 7
Date Tue, 08 Mar 2011 10:46:47 GMT
Knut Anders Hatlen <knut.hatlen@oracle.com> writes:

> Rick Hillegas <rick.hillegas@oracle.com> writes:
>
>> The error in ParameterMetaDataJdbc30Test might occur if you change the
>> order of the test cases. I believe that that order is platform
>> dependent and just happens to be alphabetical on most of the platforms
>> we test. Have we changed the version of Java 7 which we are using in
>> our regression tests?
>
> Hi Rick,
>
> The Java 7 installation on the regression test machines was updated from
> 1.7.0-ea-b116 to 1.7.0-ea-b131 on the day these failures started
> happening. I haven't been able to reproduce the failures in my
> environment when running the tests separately. I'll start suites.All
> with the same version and see if the failures can be reproduced then.

It turns out that the ordering of methods changed in b129, probably
because of http://bugs.sun.com/view_bug.do?bug_id=6990754. The exact
order was unspecified even before that, but the methods happened to
always be returned in the same order as they were specified in the class
file.

I see that there is a bug report to get the old ordering back
(http://bugs.sun.com/view_bug.do?bug_id=7023180), but even if that's
fixed, I think our tests should be changed to work regardless of the
order in which they run. When we started testing on phoneME, many tests
had to be fixed because phoneME returns the methods in the opposite
order of what most other JVMs do.

I'll file JIRA issues for the tests I've seen fail because of this.
(It's a bit hard to reproduce the failures reliably, since the order
varies between runs even on the same platform.)

-- 
Knut Anders

Mime
View raw message