Community
Participate
Working Groups
I have been using WTP 152 with its validators on a large workspace I have. I think some builders went out of their way a bit.... I now have a workspace with a lot of markers. Whenever I start it up, it takes minutes to start with the CPU at 100% and memory quickly running out. The Eclipse splash screen shows up, but no progress is shown in its "plugin loading progress bar". After 4-5 minutes of 100% CPU Eclipse gives up. It took me quite a while to analyze the log to see why my application would not launch. It ran out of memory. I profiled with Yourkit, and I have 2 questions: 1. I see that loading the markers is one of the first things we do, and we do it on the main thread. Is that necessary? Sounds to me we can do that later on? 2. Perhaps the platform can set a maximum snap file size to recover from in case builders generate too many (bogus) entries? My markers.snap file is 39MB, which seems too big too handle. I attached the profile. I can send the snap to anyone who wants. I won't attach it to this bug directly though (too big). I made the severity of this bug "major", because it renders my Eclipse useless and makes startup really, really slow.
Created attachment 51538 [details] yourkit snapshot showing MarkerManager locking the CPU
Note that this code path only occurs during crash recovery, rather than normal startup. It's possible some of this crash recovery could be done later in a background thread. *** This bug has been marked as a duplicate of 149014 ***