| Summary: | [performance] WikiText is causing a major slow down in task editor open | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | z_Archived | Reporter: | Mik Kersten <mik.kersten> | ||||||||||
| Component: | Mylyn | Assignee: | David Green <greensopinion> | ||||||||||
| Status: | RESOLVED DUPLICATE | QA Contact: | |||||||||||
| Severity: | major | ||||||||||||
| Priority: | P1 | CC: | steffen.pingel | ||||||||||
| Version: | unspecified | ||||||||||||
| Target Milestone: | 1.1.2 | ||||||||||||
| Hardware: | PC | ||||||||||||
| OS: | Windows 7 | ||||||||||||
| Whiteboard: | |||||||||||||
| Attachments: |
|
||||||||||||
|
Description
Mik Kersten
Mik, thanks for the bug. Can you point to a bug that causes the Mylyn task editor to exhibit this behaviour? Do you have a reliable way to measure the time it takes to open the task editor? I don't have any good way of measuring that reliably. We should write performance tests for that to gather data that can be used to detect regressions. (In reply to comment #3) > We should write performance tests for that to gather data that can be used to detect regressions. Agreed try testing bug with a long description, such as bug 280726 possibly related, word wrap performance bug 168557 I've done some profiling of opening the task editor. The description part appears to account for about 10% of the total load time, of which only 4% can be attributed to WikiText. I believe that recent changes to WikiText (fix for bug 283629) may have greatly reduced the delay when opening a task editor. Created attachment 145449 [details]
profiling snapshot
Attached profiler snapshot (NetBeans)
Created attachment 145450 [details]
calltree showing timings
Created attachment 145451 [details]
profiler snapshot without wikitext
Created attachment 145452 [details]
screenshot of calltree in profiler without wikitext
I've attached the snapshot and calltree with and without wikitext enabled, profiling opening of the same bug. While it's not very useful to compare absolute times (milliseconds), it's interesting to compare relative times (percentages). I suspect that this issue has largely been resolved by changes to WikiText with the fix for bug 283629. Mik, any chance you could repeat your testing with the latest WikiText? I recommend trying it out on a variety of bugs, including those that have large stack traces in the bug description. As some of the performance figures will be platform-dependant (due to the use of SWT), it will be good to get your feedback from a Windows platform (my testing was done on a mac.) Fixed by bug 283629 *** This bug has been marked as a duplicate of bug 283629 *** |