Community
Participate
Working Groups
Need a test to verify what happens if calling the refresh local operation in a resource placed in a network drive that went down. The resource should not be deleted and the user should be warned about it. Does it have a different behaviour in Windows and Linux? NOTES: JM (06/05/2001 12:34:21 PM) Moving to RTP for consideration
I have done some manual tests (since it would be difficult to automate such tests) on Linux (NFS) and Windows XP (Netware). Steps: 1 - add linked folder/file pointing to a NFS/Netware directory/file; 2 - disconnect the network cable 3 - refresh the project Result on Linux: the refresh operation blocks trying to access non-available directories/files. The operation can be cancelled though. 10-20s after reconnecting the network cable, the operation completes successfully. In both cases, no errors are presented to the user. The same thing happens when trying to access those non-available locations using other applications. Result on Windows XP: the refresh operation faces a short delay trying to access non-available directories/files, and then proceeds. The operation completes successfully (no errors are presented), the linked resources cannot be accessed and are properly decorated as being broken (the children are lost, and so are their meta data). An way to distinguish if a directory is empty or there is I/O errors accessing it is checking if java.io.File#list returns null. We could mark such nodes (and its children) as problematic and not discard them when refreshing. We could have a different flag to specify if this should be done or the nodes should be discarded during a refresh local operation. But do we want to provide such feature? It seems that nobody else (except us) has asked for it.
Closing. Unless someone (other than Rodrigo) complains, our current behaviour seems acceptable (and difficult to test automatically).