NPE in com.microsoft.gctoolkit.parser.UnifiedGenerationalParser#remarkPhase

[dsgrieve]

concurrentCyclePauseEvent is null
NPE is swallowed by try-catch in com.microsoft.gctoolkit.parser.UnifiedGenerationalParser#process method.

Reproduces with gclogs/unified/parallel/parallelgc.log

[sumitsawant]

@karianna / @dsgrieve Can I look into this ?

[brunoborges]

@sumitsawant go for it!

I'd guess the first step would be to create a unit test that exposes the issue.

[sumitsawant]

Thanks @brunoborges , I'll take a look . Can you assign this to me ?

[dsgrieve]

@sumitsawant - About a week ago, @kirk-microsoft and I traced through this issue. The issue is more complex than a simple NPE and @kirk-microsoft is working on a fix.

[sumitsawant]

@dsgrieve Ah okay ! Cool , I'll take look at something else .

[dsgrieve]

I believe that Issue #115 is tangled up in this.

[brunoborges]

/relevance-check

[github-actions]

Relevance Assessment: Still Relevant

• Summary: The NPE bug in remarkPhase() has not been fixed — the method still calls concurrentCyclePauseEvent.addCMSRemarkPhase() without a null check (line 339 of UnifiedGenerationalParser.java), and the exception is silently swallowed by applyRule().

• Evidence:

  • The remarkPhase() method still directly dereferences concurrentCyclePauseEvent without a null guard — unlike remark() (line 326), initialMark() (line 288), and cpuBreakout() (line 459), which all have null checks.
  • The related issue UnifiedGenerationalParser does not create ForwardReference for Parallel file #457 (closed via PR Fix #457 - create missing pauseEvent and publish when appropriate. #461) fixed a different NPE in the same class (pauseEvent being null in youngDetails/fullGCSummary for Unified-Parallel files without detail lines). That fix did not address the concurrentCyclePauseEvent null issue in remarkPhase.
  • The original reproduction file gclogs/unified/parallel/parallelgc.log no longer exists in the repository, making it harder to reproduce, but the vulnerable code path remains.
  • No PRs have been submitted to fix this specific bug. The comment from 2021 mentioning @kirk-microsoft working on a fix never resulted in a merged change.
  • Only one commit has touched this file (7e86099 — a workflow fix via Fix workflow #487), and it did not address the NPE.

• Recommendation: ✅ Keep open — the bug is still present in the current code. The remarkPhase() method needs a null check on concurrentCyclePauseEvent consistent with the defensive patterns used elsewhere in the class. A new reproduction log file may need to be sourced since the original was removed.

Generated by Relevance Check Agent for issue #119