Description of problem: CLI command updateBackingContent(fileName) for a resource does not persist correctly provided files. How reproducible: Every time Steps to Reproduce: 1. Create four versions of a WAR files to upload to a resource. 2. Use resource.updateBackingContent(fileName) consecutively to upload the versions of the file. Actual results: The WAR files are not replaced correctly with new content on the third and subsequent attempts. Expected results: The WAR file should be updated correctly each time updateBackingContent is invoked. Additional info: This problem is exclusive only to the CLI; the similar UI functionality works as expected. A code inspection uncovered signification implementation differences between UI and CLI. The solution is to update the CLI code to use the same APIs as the UI. Such update requires code changes in the CLI, server, and remote interfaces.
After further investigation this issue affects both the UI and CLI for deploying packages to content type resources. The CLI problem is more visible because the code relies on the existing version of a package to find the version for the current deployment. The UI has a less visible/obvious problem because the user is required to type the version of the package to be deployed and thus avoids retrieval of inconsistent data (see explanation below). The current deployment request life-cycle request is very complex; the big picture is this: 1) A package deployment is requested by the user. 2) The new packages are stored in the database. 3) Based on database information a request is generated for the agent 4) The agent receives the request and: a) copies the new packages to the correct location and sends the result of the deployment action back to the server. b) starts discovery processes for content, service and server; each of thse processes will then send their results back to the server. 5) The server receives the deployment response and does some database updates. 6) The server receives the discovery response and does some database updates. The root cause is missing/incorrect installed package information on the server. The server actions from 5 and 6 operate over the same data but they update slightly different content. The defect stems from the fact that steps 5 and 6 leave the server data inconsistent with the actual agent-side state. The issue is even more visible on Tomcat 6 because of unimplemented SHA computation functionality. Based on the current investigation code fixes are required only in the backend, agent, and plugins. This would resolve both the UI and CLI issues without explicit updates.
For verification please execute and refer to test cases for dependent bugs (in particular bug 761593, bug 767247, and bug 767393).
master branch commits: http://git.fedorahosted.org/git/?p=rhq/rhq.git;a=commit;h=0db85189e02048786d8cd6ad83d8c4ef04076846 http://git.fedorahosted.org/git/?p=rhq/rhq.git;a=commit;h=a49908fba6460d7d310268c5f187e94f7ce1614d http://git.fedorahosted.org/git/?p=rhq/rhq.git;a=commit;h=5fced2929b4840cf01b6eeab7dc35ed38b379e9b http://git.fedorahosted.org/git/?p=rhq/rhq.git;a=commit;h=1dd4054df01aa17977498813f8b9dd5fcf3bc697 http://git.fedorahosted.org/git/?p=rhq/rhq.git;a=commit;h=0800e7dfb264d2c97e7f9638e98fcad6deeeab4a http://git.fedorahosted.org/git/?p=rhq/rhq.git;a=commit;h=5d744ed0ed9cae0923f0d74e450feddc8bb2cb40
Bulk closing of items that are on_qa and in old RHQ releases, which are out for a long time and where the issue has not been re-opened since.