"Concurrent phase not completely recorded" when processing CMS Unified log

[jlittle-ptc]

Describe the bug
When parsing the attached CMS log using the unified logger, no events are being recorded. Parser logging just reports a number of warnings: "Concurrent phase not completely recorded: 0"

Note that this may be a simple vs detailed logging issue: The attached log file does not contain any CPU lines for publishing to trigger off of. (I'll update this issue if I get logger parameters used)

To Reproduce
Parse the attached file.

Expected behavior
Log is parsed correctly and events are generated.

Affected File
gc_cms_no_events.zip

[jlittle-ptc]

Flags in use: -Xlog:gc:file=/gc.log:time,uptime,level,tags:filecount=10,filesize=100m

[brunoborges]

/relevance-check

[github-actions]

Relevance Assessment: Still Relevant

• Summary: This bug is still present in the current codebase. CMS concurrent events are never published when the GC log lacks CPU summary lines, which is the case with simple unified logging flags like -Xlog:gc.

• Evidence:

  • The warning message "Concurrent phase not completely recorded" still exists at UnifiedGenerationalParser.java:301.
  • The concurrentEvent is only published inside cpuBreakout() (lines 463–469), which triggers only when a CPU summary line is matched. Without CPU lines, the event is silently dropped.
  • A closely related fix was already applied for issue UnifiedGenerationalParser does not create ForwardReference for Parallel file #457 in PR Fix #457 - create missing pauseEvent and publish when appropriate. #461 (merged Nov 2025), which added fallback publishing in fullGCSummary() and youngDetails() when !diary.isPrintCPUTimes(). However, the same fix was not extended to concurrentEvent or concurrentCyclePauseEvent, leaving CMS concurrent phases and pause events (Initial Mark, Remark) unpublished in no-CPU-times logs.
  • The concurrentPhaseEnd() method (line 312) only sets the duration but never publishes — it relies entirely on a subsequent CPU line to trigger publication.
  • No open PRs or recent commits address this issue. The most recent commits are dependency bumps and CI workflow changes.
  • The reporter confirmed the JVM flags in use (-Xlog:gc:file=(path)/gc.log:time,uptime,level,tags) — this is simple GC logging that does not produce CPU breakdown lines.

• Recommendation: ✅ Keep open — the item is still valid and actionable. The fix pattern from UnifiedGenerationalParser does not create ForwardReference for Parallel file #457/Fix #457 - create missing pauseEvent and publish when appropriate. #461 (publishing events when !diary.isPrintCPUTimes()) should be extended to concurrentEvent in concurrentPhaseEnd() and to concurrentCyclePauseEvent in remarkSummary()/initialMarkSummary() to handle GC logs without CPU summary lines.

Generated by Relevance Check Agent for issue #477