Chromium scoped: five C++ projects clear the rule, and no split of them meets the criteria¶

datasets/results/host-scoping-chromium.json, written by scripts/host_scoping.py from the raw counts committed beside it; about 1,050 requests at one a second, nothing written into the corpus. The selection rule (host-scoping-chromium/selection-rule.txt) was written at 21:58 EDT, before any per-project yield was read: C++ the dominant file type among what sampled human changes touch, and at least 256 projected examples over the train and dev months (2024-11 to 2025-10, cutoff 2024-11-01). Projected examples are human changes times reviewer-anchored comments per sampled change times 0.42, which is Qt's 0.219 examples per fetched change over those months (its examples and fetch records on disk) against its 0.5 comments a change of 2026-09-21. The 0.5 comes from ten changes, so the projections give orders of magnitude and nothing finer.

The other eight candidates are not C++ and the rule stops there: devtools-frontend (TypeScript), catapult and depot_tools (Python), luci-go (Go and TypeScript), crosvm (Rust), and the ChromiumOS C trees (kernel, ec, depthcharge). A two-week tally of the rest of the host found only small C++ projects (libyuv, libchrome, breakpad, under 30 human changes a month each). chromium/src's row averages three months (2024-11, 2025-04, 2025-10) counted in sub-month ranges; the others count all twelve. Human here means an owner that is no service account or roller, which is scoping only: the pipeline filters no owners, and a roller's change yields nothing once the author filter and the anchor requirement have run. 47% of chromium/src's merged changes are a roller's.

The rule admits five projects and the split criteria admit none. Under placebo_corpus.py's greedy rule, chromium/src takes a half alone (1 project, 100% of its half's 55,453) and the other four make 4,424 with v8 at 54%. Without chromium/src, v8 takes a half alone and the other half is three projects at 2,037, short of OpenStack's smaller half (2,163, from its frozen train split). No subset of the five passes, and admitting the C trees does not rescue it: the kernel would hold 74% of its half. chromium/src projects twenty-three times the next largest, so every set containing it fails on share, and every set without it fails on volume. A qualifying Chromium arm needs chromium/src cut below the project, by directory or component, and that is a design decision rather than a scoping result.

chromium/src cannot be fetched by month as the CLI stands. The host serves at most 10,000 results for one query and then drops _more_changes instead of refusing: 2024-11 ended at exactly 10,000, the last updated on the 13^th, and a probe past it returns Cannot go beyond page 100. A fetch would have written a month missing its first half and reported success. fetch_changes now asks for anything the query matches at or before the oldest second it received, reading past every change it was already served at that second, and raises on one it was not (checked live: v8/v8 2025-10 passes, chromium/src 2024-11 raises; the same-second case is tested offline). Collecting chromium/src needs sub-month queries merged into one snapshot, which waits on the directory decision.

The collection would be the largest the study has made. From the same table: the five projects merge about 237,000 changes over the window, one comments request each, and about 169,000 diff requests follow from their reviewer comments, about 410,000 requests or nearly five days at one a second. Without chromium/src it is about 28,000, under eight hours. The design's pacing docstring treats 70,000 as a full corpus.

robots.txt, read the same day. chromium-review serves User-Agent: * / Disallow: /, as android-review does; codereview.qt-project.org serves Disallow: / with Crawl-Delay: 3, and review.opendev.org Crawl-delay: 2. The pipeline paces all four at one request a second. What this means for collection is recorded in the re-registration entry of 2026-09-22, not here.

One matching trap for later analyses. Chromium writes C++ as .cc and Qt as .cpp; they share only .h. The separability probe and --content restrict by suffix, so a Chromium-Qt C++ cell would match headers alone unless suffixes are mapped to a language first.

Chromium is added to GERRIT and scripts/fetch_chromium.sh is the resumable driver. It takes the project set explicitly and refuses to resume a month fetched under a different set. No collection has started: there is no qualifying project set, and chromium/src cannot be fetched by month. All 1,050 scoping requests were answered, from a workstation.