Attached is a replacement for the Parallel task which fixes this behavior. You'll also need to patch the Execute task's waitFor(Process) method as follows:
protected void waitFor(Process process) {
try {
process.waitFor();
setExitValue(process.exitValue());
} catch (InterruptedException e) {
process.destroy();
}
}
If you don't feel good about just wholesale replacement of the Parallel task
(we've been using it non stop in our regressions for quite some time now), you
can just rename the class and use it as an external task.
----- Original Message -----
From: "Aaron Siri" <[EMAIL PROTECTED]>
> I'm still learning Ant and ran into a problem for which I can't see an
> obvious (elegant?) solution.
>
> The documentation for the parallel and sequential tasks present the
> following example:
>
> <parallel>
> <wlrun ...>
> <sequential>
> <sleep seconds="30"/>
> <junit ...>
> <wlstop/>
> </sequential>
> </parallel>
>
> The problem I see with the example above is that if the <junit ...> task
> fails then <wlstop/> never gets called which means that <wlrun ...> will
> never terminate so the <parallel> (and the rest of the build) never
> finishes. Is this right? If so, how would one get <wlstop/> to execute no
> matter what <junit ...> does (assuming <junit ...> is an atomic task) but
> still have the sequential report failure?
>
> An actual build script that exhibits this behavior follows. Assume that the
> <waitfor> task is a server process that doesn't time out, that the <fail>
> task is a unit test against the server that failed, and that the <property>
> task is what is used to bring down the server.
>
> <project name="mytest" default="test" basedir=".">
> <target name="test">
> <parallel>
> <waitfor maxwait="100" maxwaitunit="hour"> <!--
> start the server -->
> <isset property="stop"/>
> </waitfor>
> <sequential>
> <fail message="Oh no!"/> <!-- failing unit
> test -->
> <property name="stop" value="true"/> <!--
> stop the server -->
> </sequential>
> </parallel>
> </target>
> </project>
>
> The above will hang indefinitely (well, for 100 hours).
>
> Perhaps I'm missing something obvious or there is some built-in task that
> handles this sort of thing. Otherwise, could sequential be tweaked to add
> an attribute that would allow it to keep running tasks even if one fails but
> still return failure?
>
> Any insights or suggestions are appreciated.
>
> -Aaron
>
> --
> To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
>
>
>
Parallel.java
Description: Binary data
-- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
