Hi,
I'm using the gmaven plugin to execute a groovy script and whenever I run
release:prepare the scripts lines get joined. I use version 2.4.1 of the
release plugin, and the following snippet is affected:
<plugin>
<groupId>org.codehaus.gmaven</groupId>
<artifactId>gmaven-plugin</artifactId>
<version>1.4</version>
<executions>
<execution>
<id>kick-managed-nodes</id>
<phase>install</phase>
<goals>
<goal>execute</goal>
</goals>
</execution>
</executions>
<configuration>
<source><![CDATA[
import java.net.Socket
def s = new Socket("localhost", 80);
s.setSoTimeout(10000);
s.withStreams { input, output ->
output << "kick\r\n"
output.flush()
}
]]></source>
</configuration>
</plugin>
After executing release:prepare, the pom.xml is rewritten and the script
above get modified as:
<source><![CDATA[
import java.net.Socket
def s = new Socket("localhost", 80);
s.setSoTimeout(10000);
s.withStreams { input, output ->
output << "kick\r\n" output.flush()
}
]]></source>
The above will break, as there is no separation between 'output <<
"kick\r\n"' and 'output.flush()'. This can be solved by adding semicolons
at the end of the lines in the script however I think the release plugin
should not reformat anything inside a CDATA block. There is an existing
open issue regarding this:
http://jira.codehaus.org/browse/MRELEASE-601unfortunatelly I was not
able to add new comment to it.
The problem as I understand is that the release plugin first removes all
extra whitepaces as they would cause problem with JDOM, and this is the
step that causes the CDATA section to be altered. The matching part is:
output << "kick\r\n" output.flush()
}
]]>
I don't see an easy way to deal with excluding CDATA from the applied
regexp replacement, maybe someone on the list does.
Thanks,
Andras
