Bugzilla – Full Text Bug Listing
| Summary: | [perfs] SearchAllTypeNames performance tests are slower on eplnx2 | ||
|---|---|---|---|
| Product: | [Eclipse Project] JDT | Reporter: | Frederic Fusier <frederic_fusier> |
| Component: | Core | Assignee: | Frederic Fusier <frederic_fusier> |
| Status: | VERIFIED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | P3 | CC: | frederic_fusier, kim.moir |
| Version: | 3.5 | ||
| Target Milestone: | 3.5 M4 | ||
| Hardware: | PC | ||
| OS: | Windows XP | ||
| Whiteboard: | |||
|
Description
Frederic Fusier
Problem comes from the perf_34x branch which was not versionned, hence the map file was not updated... So, the code used to run performance tests for baselines is the v_874 one! Unfortunately, v_874 had an issue while setting the JRE_LIB classpath variable with JDK 6.0. This was fixed in both streams but not taken into account for baselines as the JDT/Core map file didn't change in that branch... So, tests run with a valid JRE_LIB in HEAD but without it in perf_34x which explains the delta between the values for search all type names tests (more than 8,000 types more found in HEAD...). Note that it also explains why other test boxes do no show this delta as they are running perf tests with a JDK 1.5 which works properly without any changes needed... org.eclipse.jdt.core.performance.tests perf_34x contents has been synchronized with HEAD contents. It has been versionned and the map file has been updated. So, consider this released for 3.5M4 Kim, Do the baseline perf tests run each Friday for perf_34x branch from now? Yes, I'll update the build schedule to reflect this.... Thanks :-) Verified for 3.5M4 using I20081209-0100 |