Community
Participate
Working Groups
There is ~10% regression in this test (org.eclipse.core.tests.runtime.perf.StartupTest#testApplicationStartup) starting 6th or 7th May. The regression is not really clear. It is seen both on linux and windows. Unfortunately, the windows performance results are not generated, but I could see it in the DB.
That's the headless startup test case.
I'm not seeing the issue start with the 6th or 7th build. The regression seems consistent before the 6th or 7th. My suspicion is that this test is invalid because each release we add more bundles and that contributes to the overall startup time. Headless starts make this more evident because the overhead of starting the framework is pretty much the whole test.
Do you plan to do something with this bug?
(In reply to comment #3) > Do you plan to do something with this bug? No, the slowdown can be attributed to the growing number of bundles.
Bug is getting to too, we have to look again once the performance test are running again.