Reported by an AccuRev plugin user on the users list:
–
We are using Hudson to perform our builds against an Accurev server and are getting failed authentication errors in the middle of the build. It had previously authenticated successfully and issued successful Accurev commands so it seems like something is stealing the login.

I've tried setting an ACCUREV_HOME=C:\ environment variable but Hudson seems to ignore that and continues to use the session file in the project's job directory.

I believe there are 3 separate processes that want to authenticate with Accurev, Hudson itself for polling and checking for changes, the individual job for populating the build workspace and then we run our builds with nant and it too issues accurev commands. If I run one build at a time they are all successful.

I've checked on the checkbox that says to serialize the Accurev commands and that seemed to have solved a different problem we were having.

Has anyone else experienced this problem? Does anyone have any suggestions?
–

There does seem to be a bug here, and there are potentially a pair of issues:

1. The auth failed issue
2. The ACCUREV_HOME env variable is being ignored in favor of the job directory.

The latter may be as designed. The code explicitly sets ACCUREV_HOME to the parent dir of the workspace dir on the building node, regardless of if a value is already set. Interestingly, if the job runs on a slave node, then that directory is the workspace dir shared by all jobs on the node, whereas on the master it is the job dir for a particular job. I'm not sure why it shares credentials on a slave but not the master. This may be a separate bug, and if I decide that it is as part of researching this issue, I'll open a separate issue to track it. If it is the cause of this problem, then this issue will cover it.

Next step is to get an interim build that logs the CLI output in the event of any failure so I can find out exactly it is saying. Currently it only logs the return code.