Support gc logs where each gc event is described on a single line

[AlinaHovakimyan]

GCtoolkit does not support simple logs. That is, logs where each GC event is described on a single line. Such log, for example, can be generated, when using Parallel GC with unified logging and minimal arguments.
Example vm args: -XX:+UseParallelGC -Xlog:gc:file=gc.log
Example log: gc_parallel.log

Why this is important:
These minimal-configuration logs sometimes are the first logs users produce when diagnosing GC behavior, before they move on to more advanced logging options.

[karianna]

Hi AlinaHovakimyan - that is somewhat by design in terms of the scope from the original authors (we typically guide users through the minimal useful settings for GC log analysis). We're not saying that we would never support these simpler logs, but I think we'd be looking for community contributions there and making sure that the end use is aware that they should really be adding a minimal set of flags.

[dsgrieve]

Martijn Verburg (@karianna) - this looks like issue #372, or at least the reason why gctoolkit needs more than just default logging.

[AlinaHovakimyan]

Thank you Martijn Verburg (@karianna) and David Grieve (@dsgrieve) for your reply. Do you know if adding this support will require significant changes ?

[karianna]

Thank you @karianna and @dsgrieve for your reply. Do you know if adding this support will require significant changes ?

I'm not sure, I suspect hte parser work will be fairly minimal, but surfacing the data may be more complex as the domain layers expect richer data.

[jlittle-ptc]

Also #457 is a similar issue which I've submitted a fix for, but is not yet available in release.

[brunoborges]

/relevance-check

[github-actions]

Relevance Assessment: Likely Outdated

• Summary: The core problem described in this issue — parsing single-line Parallel GC events from minimal -Xlog:gc logs — has been fixed by PR Fix #457 - create missing pauseEvent and publish when appropriate. #461 (merged Nov 17, 2025), which resolved the closely related issue UnifiedGenerationalParser does not create ForwardReference for Parallel file #457. The fix is present on main but has not yet been included in a release.

• Evidence:

  • PR Fix #457 - create missing pauseEvent and publish when appropriate. #461 (merged) directly addresses this scenario. It modifies UnifiedGenerationalParser.java to create and publish pauseEvent when processing youngDetails or fullGCSummary lines without prior header lines — exactly the single-line event format described here.
  • A test case (testParallelNoDetails in UnifiedGenerationalEventsTest.java) was added validating parsing of logs like:

    [1.887s][info][gc] GC(0) Pause Young (Metadata GC Threshold) 516M->42M(7065M) 14.443ms
    

  • Issue YoungDetails event not published for GC log collected with -Xlog:gc (no *) #372 (G1 equivalent, closed) was separately fixed in UnifiedG1GCParser.java with similar logic for -Xlog:gc logs.
  • The latest release is gctoolkit-3.7.0 (May 2025), which predates PR Fix #457 - create missing pauseEvent and publish when appropriate. #461. The fix is not yet released.
  • Known remaining edge case (noted in PR Fix #457 - create missing pauseEvent and publish when appropriate. #461): Logs without the Using Parallel preamble line (e.g., rolled-over log files) cannot be uniquely identified as Parallel GC because the single-line event format collides with Serial GC. This is a diary/collector-identification limitation, not a parsing limitation.

• Recommendation: 🗄️ Consider closing — The primary parsing problem is resolved on main by PR Fix #457 - create missing pauseEvent and publish when appropriate. #461. This issue was filed 11 days after that fix was merged, likely without awareness of it. The remaining edge case (collector identification without Using Parallel) is a separate, narrower concern that could be tracked in its own issue if needed. A new release would make the fix available to users.

Generated by Relevance Check Agent for issue #472

[brunoborges]

Martijn Verburg (@karianna) consider fixing this one