The project at my company does this.
<junit fork="yes" forkmode="once" printsummary="no" haltonfailure="no">
The forkmode="once" is important for speed.
Ronald.
On Thu Feb 14 18:33:23 CET 2008 Velocity Developers List
<[email protected]> wrote:
All,
While updating my unit test for VELTOOLS-103
(https://issues.apache.org/jira/browse/VELTOOLS-103), found that I
needed to change the "fork" attribute in the <junit> directives in the
ant scripts to "true" (they are currently "false", indicating that the
unit tests will be run in the same JVM as ant).
I was wondering if anyone could explain why this is necessary. The
classes I'm trying to use were clearly being loaded correctly (as they
were running properly, with the exception of the fact that they were
behaving improperly unless fork="true") so it's not a classpath issue.
In this particular case, the library in question is attempting to change
the value of the "java.protocol.handler.pkgs" system property. I
checked, and the system property is being properly modified. Maybe the
problem is that the JVM's URL handling subsystem had already been
initialized before I changed the value of this system property, and my
URL handler was therefore being ignored (which is the problem I get when
fork="false").
Anyhow, I'd like to understand what's going on only secondarily.
Primarily, I'd like to know if it would be okay to set fork="true" for
Velocity Tools' unit tests in order to support a test I'd like to add.
Thanks,
-chris
