[Clear1nsight]

Change and mathematical scope

The primary baseline owner currently tries to install (1 GiB, 1 GiB) even under a lower inherited hard address-space limit, causing setup to fail before runner launch. This source successor reuses the existing inherited-limit calculation, forwards the legal tuple to the actual child callback, records requested/inherited/installed limits separately, and rejects parent-limit drift.

The original requested 1 GiB upper bound, scientific driver, stage/source pins, error ordering and admission gates remain intact. This is a scoped source repair with executable pure projection controls. It does not establish workload fit, complete caller qualification, a speedup, scientific acceptance or execution permission.

Task and overlap

Origin: PR #6321. Immediate baseline source: PR #6837, head 92bd2213be787cc4451a3b61b00793dad26bebdd. Reused calculation: PR #6830, head 39fcd67c991bd72918080b1d2c02feb3d679206d. Base: 7c49fe32c55de02c2c873655c804b16e4c3fef0a.

The package is research/h25-inherited-baseline-limits. It retains a byte-reversible source diff and exact provenance. The inherited-limit body stays in the baseline closure to avoid importing candidate bootstrap code.

The donor explicitly selected a common-supervisor amendment; its delta is documented separately in the package. The full execution policy remains an external pinned prerequisite. The common supervisor, stopped-entry/query partition, receiver, retention and settlement join still require connected implementation and independent review. No silent claim of compatibility with the earlier restricted protocol is made.

Validation

The exact source passed 7 changed-boundary tests in normal Python and 7 with -O, under a bounded local guard. Each mode includes 440 feasible-set oracle cases, lower inherited/zero bounds, invalid types and ordering, actual child-callback forwarding, distinct launch fields and parent-limit drift. Raw unittest results and a bounded package integrity check are retained.

Cold reproduction from the package directory:

python3 -B verify_source.py
python3 -S -s -P -B validation/test_baseline_as_successor.py
python3 -O -S -s -P -B validation/test_baseline_as_successor.py

The tests execute selected production definitions and actual call-site projections using the original pure schema predicate and an inert setrlimit recorder. They do not import/execute the baseline or candidate modules, mutate native limits through the selected callback, launch science, rerun old controls or supply full runtime qualification. Hosted archival planning is separate from these local results.

Credit

Retains the CrocSwap project and predecessor contributions in #6321, #6837 and #6830, with exact commits in PROVENANCE.json. Prepared under the Clear1nsight contribution with substantial OpenAI Codex and ChatGPT Pro assistance. Apache-2.0; existing project notices are retained.

Open next questions

Adopt the successor into the exact authenticated baseline owner/receiver closure and review its actual caller with the common-supervisor changes. Preserve original refusal/startup controls from #6818, native-entry controls from #6823, and reuse the complete-outcome CLOSE source in #6762. A complete associated environment, numerical wall/CPU/RSS bill, actual funding and original prerequisite outcomes remain necessary before execution.

[Clear1nsight]

Verification for published commit efb116491184bd16c5d1ada60d71efdefee61691: hosted run 42364922738, attempt 1, completed successfully with plan successful and replay skipped. Skipped replay supplies no scientific execution result.

Original-investigation pointer · Research-hub pointer. The PR description and README retain the source/control scope and remaining integration requirements.