Community
Participate
Working Groups
After upgrading to 3.4M7 (from 3.3) the rename refactoring seems to be quite unstable. Typically every second big rename refactoring stops in the middle of the process and leaves duplicate classes (before and after). This happens both for class or package refactoring. The problems are not really deterministic, the occur not every time. After replacing our code from HEAD and retrying suddenly the refactoring works. Maybe a concurreny problem? Attached several exception messages. (I'm not sure whether this happens only if there is a compile error in one of the refactored files.)
Created attachment 99664 [details] Error log with stacktraces
Created attachment 99665 [details] Eclipse Project that caused the problem Not: this is a maven project. You need to run a mvn eclipse:eclipse to get the required dependencies.
I attached the project where it happens. What I tried is to do a rename of the package org.hudson.pmd to org.hudson.checkstyle in any of the 4 source folders (recursive package rename). Sometimes it breaks in the src/main, sometimes in the src/test folder.
It sounds like some external process is keeping a file handle on the resources being moved. Does it happen after a reboot? If it still happens, can you please make sure that: - the attachment is not marked as text/plain - the project doesn't depend on Maven
Well, I think Eclipse holds these file locks. I even can't remove these zombie files in Eclipse: I need to exit and restart Eclipse, then the locks are removed and I can remove them again in Eclipse. There is no reboot required.
Just one additional remark: I'm using the subversive team provider.
(In reply to comment #4) > If it still happens, can you please make sure that: > - the attachment is not marked as text/plain > - the project doesn't depend on Maven
Also does it happen in a plain Eclipse SDK, or do you have other plugins installed?
I'll wait for the RC1 and re-test. There are some other save bugs already fixed, maybe these correlate...
I installed 3.4RC1 and the bug seems to be fixed here. I set the resolution to WORKSFORME...
Setting target milestone as per JDT/Core process
marking VERIFIED (verified by the bug reporter)