Chromium's V8 pool, read within its grant, cannot cover a matched draw¶
# research(2026-10). The pool fixed in the entry above, collected at 64a059c
(datasets/logs/chromium-v8-pool-2026-10-16/, raw snapshots in datasets/gerrit-granted/chromium,
backed up):
| stage | v8/v8, 2024-11 to 2025-08 |
|---|---|
| merged changes listed, created from 2024-11-01 | 7,885 |
skipped: the listing records no comment (uncommented_by_count) |
3,683 |
skipped: past the 170-a-month cap (over_read_cap) |
2,502 |
| changes read | 1,700 |
| examples built | 756 |
| after refinement (71 the reviewer's own suggested edit) | 685 |
| after deduplication (13 exact, 13 near copies) | 659 |
The comment-count probe held (comment-count-probe-chromium.json: none of 300 zero-count rows
carried a comment, every control did, upper 95% bound 0.0099), so the 1,700 reads plus its 330
are the grant's "about 2,000". V8's inline comments are mostly not a reviewer's on code a later
patch set changed: of the comments the build read, 4,441 sat on the commit message or patch set,
2,818 were the author's own and 1,242 lay outside any hunk the next patch set changed.
A matched foreign draw is N = 1,850 examples, so 659 cannot cover one, and at this yield the whole 2,000 changes would give about 780. By item 5 of the pool rule the pool is not widened past the grant, and by item 1 of the H2 plan the source falls to the next of Fuchsia and AOSP whose maintainers grant written permission before the first H2 development contrast, in that order. Fuchsia has not answered. AJ reads Chromium's infra-dev answer as covering android-review (entry above), so AOSP is the source unless Fuchsia grants first, with its projects and window as item 4 of the H2 plan fixes them. The V8 pool stays held, enters no figure, and is reported as the source tried first.