Community
Participate
Working Groups
While verifying perf results for I20090331-0901 build, I saw that some JDT/Text 'Delete *' tests do not last enough (<100ms). Here are the numbers on RHEL 5.0 machine: - DocumentPerformanceTest#measureDeleteInsert(): 18ms - DocumentPerformanceTest#measureInsertAtEnd(): 33ms - DocumentPerformanceTest#measureInsertAtStart():76ms
We have them so that we know when they go through the roof.
(In reply to comment #1) > We have them so that we know when they go through the roof. > Ok, so it could be interesting to think about a mechanism in the framework to identify that such test duration is expected... And also report a warning otherwise...
This is already possible: we could specify an absolute band but so far I didn't see the real benefit given bug 89804 is marked as WONTFIX.
(In reply to comment #1) > We have them so that we know when they go through the roof. > Reopen as I think then this test should not be put in the fingerprints. One of the goal of fingerprint is to show tests which time is meaningful for users. Other tests should be only visible in the status table.
Sorry, but even if the number is below 100 it can interest the user whether it became (much) slower or faster. And document manipulation is central to Text. Also, I agree the fingerprints are targeted to the user but let's be serious: the main consumer/user are we.
Verified for 3.5M7