Community
Participate
Working Groups
By chance, I came across a method in JavaModelListener in org.eclipse.jem.workbench.utility named protected boolean isInClasspath(IJavaProject testProject, List someJavaProjects, boolean isFirstLevel, Set visited) Not sure about earlier version, but in 3.7 M6 the compiler flags i++ in following 'for' statement as "dead code". I thought this was a error in compiler, but on second look, I think it is deadcode, since the 'return' is in the loop, so 'i' would never be incremented. In other words, the method only ever checks 1 java project, instead of the whole list. This seems like a bad bug, if the method is used (and, does seem to be used). for (int i = 0; i < size; i++) { javaProj = (IJavaProject) someJavaProjects.get(i); return isInClasspath(testProject, javaProj, isFirstLevel, visited); }
Created attachment 192184 [details] patch to implement presumed intent Assuming the intent is to look at whole list of Java Projects, this patch implements the for loop correctly. If the intent is really to just look at first Java Project in list, then code should be modified to note use a loop, since that's pretty confusing.
I went ahead and released to head. If you think important, you might want to consider "porting" this back to maintenance stream, but I know it has been running this (incorrect) way for a long time, so may not matter? Or, even if my proposed fix is "correct", it may break some other cases just because they weren't expecting it? I incremented service field, so version is now 2.0.400. Apparently this, and null checks (in bug 341367) are only changes from maintenance. Hope you don't mind me releasing it ... I know Carl usually releases everything? ... just let me know if you want me to adhere to different procedure. Thanks,
