java.io.NotSerializableException: com.xebialabs.overthere.ssh.SshSftpFile

[xebialabs]

java.io.NotSerializableException: com.xebialabs.overthere.ssh.SshSftpFile

[hierynomus]

Why are you caching the remote file in your step/deployed? You shouldn't do that...

[bmoussaud]

because before deploying it I need to upload the file on the remote host that will process the file, so I change the referenced file (LocalFile) by the new one, an SshSftpFile.

[hierynomus]

But how is that repeatable then? You shouldn't keep the reference to the remote as a field. If the step needs to re-execute, it needs to re-upload.

[bmoussaud]

What should be repeatable? I change the reference from the localfile to the SSHFile to avoid to move the file back to the deployit server, and preferably to move it directly to remote host.

[vpartington]

The SshSftpFile is not serializable because it contains the SshConnection which is not serializable. The same connection will not be there when the server is restarted.

[vpartington]

The step should be repeatable because it might fail and then need to be reexecuted. The remote file could be deleted in the meantime. Especially if the server has crashed in the meantime.

[bmoussaud]

@vpartington: protected transient C connection;

[vpartington]

Apart from the fact that this will cause NPE's all over the place unless we check whether connection is null everywhere, this will not work because we need the connection to access the file. Without a connection, the file is meaningless.

If you want to get at a file at a later stage, save the path and the connection details and then access it again. Jeroen's scenario has the advantage of not requiring that and being repeatable.

[bmoussaud]

In my use case, the file isn't stored in the temp directory but in an explicit path.

Le 18 nov. 2011 à 13:45, Vincent Partington a écrit :

Apart from the fact that this will cause NPE's all over the place unless we check whether connection is null everywhere, this will not work because we need the connection to access the file. Without a connection, the file is meaningless.

If you want to get at a file at a later stage, save the path and the connection details and then access it again. Jeroen's scenario has the advantage of not requiring that and being repeatable.

---

Reply to this email directly or view it on GitHub:
#24 (comment)

[vpartington]

OK. So then you'll need to save just the path for later reference.

[bmoussaud]

please look at https://github.com/xebialabs/pocs/tree/master/liferay-plugin before asking me to do differently

[vpartington]

Will not implement because it adds unnecessary complexity.