shuke987 opened a new pull request, #65874:
URL: https://github.com/apache/doris/pull/65874

   ### What problem does this PR solve?
   
   
`CacheHotspotManagerTableFilterTest.testShouldWarmUpPerformanceManyRules200k` 
intermittently fails its 3-second wall-clock assertion while still producing 
the correct 19,950 matches. The same timing-only signature appeared in FE UT 
builds 995541, 999142, and 1001937 across unrelated PRs and different agents.
   
   FE UT runs Java 17 with JaCoCo coverage and 12 forked JVMs. The target was 
the first performance method in the class and measured a cold regex/stream path 
once, so JIT compilation, GC, or scheduling pauses could be charged to the 
sample. In the slowest build, four adjacent performance assertions also failed 
and the class runtime grew from about 55 seconds in a nearby pass to about 134 
seconds, without a functional mismatch.
   
   ### What is changed?
   
   - Warm the same stream/lambda/regex path with 20K names outside the timed 
section.
   - Measure three complete 200K passes and apply the original 3-second 
threshold to the best sample.
   - Keep the original 15-rule workload and assert exactly 19,950 matches in 
every measured pass.
   
   This filters cold-JIT and isolated scheduling pauses while preserving the 
absolute performance guard: a persistent slowdown still fails all samples.
   
   ### Validation
   
   - Pre-fix local baseline: 3/3 passes at 439 ms, 391 ms, and 342 ms.
   - Fixed exact test in fresh JVMs: 3/3 passes at 92 ms, 156 ms, and 94 ms.
   - Fixed exact test with JaCoCo coverage: 1/1 pass at 300 ms.
   - Latest `origin/master` worktree exact test: pass at 72 ms.
   - `git diff --check`: pass.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to