Memory pool size data

It is unclear that the parsers are collecting the right data needed to support calculations such as Memory Pool Size % based on the number of regions allocated to Eden, to, archive, reserved and old space.

[dsgrieve]

I think the parser collects the right data, but then there is a calculation to turn the number of regions into a number of bytes based on region size. The trouble is that the value for regionSize might be zero. It might be better to just set the number of regions.

[kcpeppe]

I think there is a data structure to collect the region information. I'll drop more testing in here today if that is ok with you David.

[brunoborges]

/relevance-check

[github-actions]

Relevance Assessment: Still Relevant

• Summary: While PR Heap region size for G1GC is no longer parsed after JDK-8244817 #459 (merged Nov 2025) fixed a specific parsing regression for heap region size, the broader concerns raised in this issue — about the robustness of memory pool size data collection — remain unaddressed.

• Evidence:

  • Partially fixed: PR Heap region size for G1GC is no longer parsed after JDK-8244817 #459 fixed a case-sensitivity bug where Heap Region Size (capitalized, per JDK-8244817) was not matched by the parser's pattern. The regex is now Heap [Rr]egion [Ss]ize: in UnifiedG1GCPatterns.java:57. This directly fixed one cause of regionSize being zero.
  • Still an issue: heapRegionSize is initialized to 0 in G1GCForwardReference.java:45, and if the GC log doesn't contain a Heap Region Size line at all, it remains 0. All region-to-byte conversions (Eden, Survivor, Old, Humongous, Archive) are guarded by if (heapRegionSize > 0) checks (lines 580–624), meaning the data is silently dropped rather than flagged or handled gracefully.
  • Comment suggestion not implemented: @dsgrieve's original suggestion (March 2023) to "just set the number of regions" instead of converting to bytes was never acted on. The code still relies entirely on heapRegionSize for byte conversion.
  • No duplicate or superseding issue was found — this is the only open issue covering this topic.
  • Comment from original author (@kcpeppe) mentioned they would add more testing, but no follow-up tests for region data correctness appear to have been added.

• Recommendation: ✅ Keep open — The issue's core concern remains valid. PR Heap region size for G1GC is no longer parsed after JDK-8244817 #459 was a targeted fix for one parsing regression, but the broader questions about data collection robustness (e.g., what happens when heapRegionSize is unavailable, and whether the approach of converting regions to bytes is the right design) are still open. This issue can serve as the tracking item for that work.

Generated by Relevance Check Agent for issue #285