Description Manoj N Palat 2021-03-29 10:00:13 EDT

Related to bug 551899

For the record, from https://ci.eclipse.org/jdt/job/eclipse.jdt.core-Gerrit/4337/testReport/junit/org.eclipse.jdt.core.tests.compiler.regression/GenericTypeTest/testBug543480BasedOnTest2FromComment4ToSameSameOptimization___1_8/

----------
Potential fluctuation of a performance test: average long compile time 1072.1666666666667ms should be less than 10.0x the average short compile time 6.166666666666667ms
long compile times: [#9 349ms (excluded), #10 366ms (excluded), #8 575ms, #5 1008ms, #4 1103ms, #7 1177ms, #3 1180ms, #2 1390ms, #1 1724ms (excluded), #6 9817ms (excluded)]
short compile times: [#3 5ms (excluded), #2 6ms (excluded), #4 6ms, #7 6ms, #8 6ms, #9 6ms, #10 6ms, #1 7ms, #6 7ms (excluded), #5 15ms (excluded)]
----------

Failure seems to happen more frequently (no data to back this claim though) when the load is high.

Would it mean that we need to fall back on Stephan Herrmann's suggestion of Plan B in bug 543480 comment 62 - quote : "As a plan B I would suggest to add test-only API to indeed measure the number of bounds created, even during a full compilation."