RE: [Ant-contrib-developers] Potential New Ant-Contrib tasks: A ntServer and RemoteAnt

[Inger, M. <inger@Synygy.com>]

There's an alpha version of this in cvs now.

-----Original Message-----
From: Inger, Matthew 
Sent: Thursday, November 06, 2003 4:00 PM
To: 'Scott Hebert'; Inger, Matthew;
'ant...@li...'
Subject: RE: [Ant-contrib-developers] Potential New Ant-Contrib tasks: A
ntServer and RemoteAnt


Through the right use of ThreadGroups and BuildListener, I suppose I could
track the status of which
targets get executed during the course of a command.  I can then build an
XML document,
and send this with the response.  How we would want to display these results
in the calling
buildfile, i'm not really sure yet.
 

-----Original Message-----
From: Scott Hebert [mailto:sc...@ca...]
Sent: Thursday, November 06, 2003 2:57 PM
To: 'Inger, Matthew'; 'ant...@li...'
Subject: RE: [Ant-contrib-developers] Potential New Ant-Contrib tasks: A
ntServer and RemoteAnt


Here is what I would expect to see as the response that gets sent back to
the client:
 
Buildfile: build.xml
 
BUILD FAILED
Target `target1' does not exist in this project. 
 
Total time: 1 second
 
It may nice to also see which targets have been called out by the build. Not
the target's output though. Example:
 
Buildfile: build.xml
 
[target1]

[target2] 
[target3] 
[targetN]
 
BUILD SUCCESSFUL


Total time: 25 seconds

Does this seem reasonable?
 
SH

-----Original Message-----
From: Inger, Matthew [mailto:inger@Synygy.com]
Sent: Thursday, November 06, 2003 2:41 PM
To: 'Scott Hebert'; 'ant...@li...'
Subject: RE: [Ant-contrib-developers] Potential New Ant-Contrib tasks: A
ntServer and RemoteAnt


Right now, the initial implementation just does straight socket
communication,
using ObjectInputStream and ObjectOutputStream.  The buildfiles have to
be on the remote host as well.  Trying to have the build file on the local
host
and send it over would be a tough thing to do, since you'd have to worry
about
all the dependencies (such as other .xml files, .jar files, and any external
resources).  Using SOAP seems a little complicated for what this task is
designed
to do.  I'd rather not force it to have to run a webserver.
 
Right now, you just start the ant build on the remote machine, and let it
sit
there waiting for requests.  It's a lot more lightweight.
 
The other thing i was wondering was what to send back to the client.  I've
built
in a Response object which is streamed back to the client, but right now it
only has two items, a boolean flag which indicates if the command was
successfull,
and a string which contains the error stack trace if any (it's a string due
to possible
task limitations.  BuildException is not Serializable [blame the Location
class], and
it's possible other exceptions that occur might not be).
 
Would we want other information sent back along with the response?
 
Thoughts?

-----Original Message-----
From: Scott Hebert [mailto:sc...@ca...]
Sent: Thursday, November 06, 2003 12:21 PM
To: 'ant...@li...'
Subject: RE: [Ant-contrib-developers] Potential New Ant-Contrib tasks: A
ntServer and RemoteAnt


I think this is super idea.
 
I liked some of the features of RANT (
http://sourceforge.net/projects/remoteant/
<http://sourceforge.net/projects/remoteant/> ) but development seems dead.
It made use of a SOAP service to perform the build. It was not without its
limitations though. I believe the build file had to be located on the remote
host and the output wasn't stream back to the client making the call.
 
I would vote to have to the AntServer built to communicate over HTTP to
permit implentations in firewalled environments.
 
SH

-----Original Message-----
From: Inger, Matthew [mailto:inger@Synygy.com]
Sent: Thursday, November 06, 2003 10:28 AM
To: 'ant...@li...'
Subject: [Ant-contrib-developers] Potential New Ant-Contrib tasks: AntServer
and RemoteAnt



I know there's probably one or two out there, but the ones I saw didn't
quite fit what 
I was looking for. 

The basic idea is that on a host machine, someone could run a buildfile
which contains 
a call to the <antserver> task.  This would block the host build, and the
<antserver> task 
would wait for connections on a specified port.  When a connection is
received, it would 
read commands from the socket, and execute the specified commands in the
context 
of the current build file. 

        Example: 
                 <antserver port="12345" /> 

On the client machine (which could be the same as the host machine), a
buildfile would 
contain a call to the <remoteant> task, and as subelements, would specify
what commands 
they'd like to send to the server.  it would then open a connection to the
server, and send 
the commands to be executed. 

        Example: 


	        <remoteant machine="server.machine" port="12345"> 
            <runtarget target="execute.build"> 
               <property name="build.type" value="full" /> 
            </runtarget> 
            <runant dir="tests" target="build.tests"> 
               <property name="build.type" value="full" /> 
               <reference refid="my.ref" torefid="inherited.ref" /> 
            </runtarget> 
        </remoteant> 

	would be the equivalent of running the following directly on the
server machine, from within the same buildfile where the <antserver> task
was run 

    


	        <antcall target="execute.build"> 
           <param name="build.type" value="full" /> 
        </antcall> 
        <ant dir="tests"> 
           <property name="build.type" value="full" /> 
           <reference refid="my.ref" torefid="inherited.ref" /> 
        </antcall> 

    

Does anyone have any thoughts on this?