Community
Participate
Working Groups
Using build I200405110800. Scenario is still the same than bug 55207, bug 55289...: merge JDK_1_5 branch with HEAD. There's one file which is quite big in jdt-core (223K): org.eclipse.jdt.internal.compiler.parser.Parser.java Unfortunately, this file has many changes in JDK_1_5 branch and when I need to merge it, it takes a long while to compute differences... This is really bad when I got conflict and need to look inside method as it recomputes each time I select a new method in structure pane! Last problem is that it recomputes twice when selecting a method created in JDK_1_5 and which does not exist in HEAD stream. Perhaps are there some optimizaiton in this area which can be done? Set severity to critical as I'm afraid merge would not be possible with this performance problem if during one merge I had lot of conflict in this file.
Also, one question for my personal knowledge: is comparison perfomed on server or on local box? Subsidary question: If comparison is done on server, I guess that file is sent to it, right? I want to know if it would possible to do the merge from home... Thx
Re: comment #1: The compare editor used by merge is provided by Platform Compare. Re: Comment #2: When you perform an Update in the Merge particpant, the local contents are sent to server abd the server will attempt to merge the differences into the local contents. However, we have modified the update to skip files that contain conflicts since these are not handled well by CVS. For these files we recommend using the compare editor in Eclipse. Also, we have added optimizations to the merge participant to simplify (as much as possible) the handling of ongoing merges where a user merges multiple times from one branch to another (such as in your case). The effect of this is that you can attempt the merge from another machine but will lose the history of previous merges that you did on your work machine. Moving to compare to address comment #1
I have a similar problem comparing two files both on 2.1.3 and 3.0. What I did was a ittle refactoring. The first thing I did was to reformat the entire file which is about 130k, then I manually went through the file to modify the format to my satisfaction. Later before I checked in the changes, I compared the file with the HEAD from CVS. The progress bar advanced to about a quarter, and then almost stopped there. After a while, an error message was popped up: Too many differences. Turn on the 'Igore white space' option or do a structure compare first. I turned the 'Igore white space' option on. But the result was the same. Trying to do a structure compare also showed the same result.
*** This bug has been marked as a duplicate of bug 109254 ***