44 lines
6.3 KiB
Markdown
44 lines
6.3 KiB
Markdown
# Independent text-producer runtime evidence review
|
|
|
|
Disposition: final source `97a8703adc75467a7f8e75f5c6cb00ca542b707c` satisfies the bounded regression comparison, with the explicitly documented single framework-hint difference below. This is not a wholly passing suite: full core tests retain 16 failures, and workspace format retains exit 1. No Cargo, test executable, debugger, archived evaluator, or network operation was run during this review.
|
|
|
|
## Reproducible readback
|
|
|
|
Ran the reviewed local helper using bundled Python with `-I -B`, arguments `97a8703adc75467a7f8e75f5c6cb00ca542b707c 97a8703 6`. Current `inspect_runtime.py` SHA256 is `ea4d783dc2d791b4b6f8dae105620fbd216d49b859af3123957d22d7de1ca2d5`. Its exclusive result `TEXT-PRODUCER-RUNTIME-READBACK.json` SHA256 is `2e15ae1800b4d01cb9f854f362a23a2d25fcabd8d77f39c0ce2001c717d535ed`. It completed with exit 0.
|
|
|
|
The original stricter checker SHA256 `bdf2bb33dd54916440b8dfde1b8bd38d61d3975d8c760deeccf1c92eb562c8d3` is preserved unchanged as `inspect_runtime_strict_original.py`. Its genuine refusal and the precise amendment are retained in `CHECKER-NOTE.md`; no source or test was changed to satisfy this comparison.
|
|
|
|
The baseline is source `8d52dab97c92cc31baae9f20d8c324b55818aa5d`. Baseline full/list/fmt logs and receipts match the local committed emission evidence ZIP byte-for-byte, archive SHA256 `8d46b31379d75567b195b639cb3bc289ad1c9d26df53d29dbfd99ba22e38c5b9`. This review does not claim remote publication of that archive.
|
|
|
|
## Actual final gates and identity
|
|
|
|
| Gate | Exit | Actual result |
|
|
| --- | ---: | --- |
|
|
| focused | 0 | Six new text-producer tests passed; 269 filtered |
|
|
| full | 101 | 259 passed, 16 failed, zero ignored; 275 unique results |
|
|
| list | 0 | 275 unique names, exactly the full runtime set |
|
|
| strict core all-target Clippy | 0 | Completed `-D warnings` |
|
|
| workspace format | 1 | Exact baseline raw output retained |
|
|
|
|
All 269 predecessor outcomes are unchanged. The focused set is exactly the six new full/list names, all passed. All ten stack-named tests in the final full log passed. There is no inference of additional fresh-process repetitions for this source.
|
|
|
|
All five gate receipts bind clean current Git HEAD, the same 556 declared input paths and raw bytes (compared against actual Git blobs and current disk), unchanged input hashes after each run, and the same captured launcher. Input manifest SHA256 is `907d453bc94eada74a42e2a3c2056660515102273283cb4867f6ab82c887e52a`; captured launcher SHA256 is `bcd83039653815c485ec18fb8d2a75dfa76a39c6e3debca6f525c35a6e474b3a`. The launcher was read, not executed: it records unfiltered merged process stdout/stderr and the actual return code, checks declared source bytes before/after, captures exact helper bytes, and derives executable paths from Cargo output. Final commands are locked/offline core library focused/full/list, locked/offline core all-target strict Clippy, and workspace format check. Single-job, nonincremental environment matches the baseline; stack-size and lifecycle overrides are absent.
|
|
|
|
Focused/full/list bind the same actual current executable: 42,614,784 bytes, SHA256 `41da1a4d788e2bd7392e69a724b7de259752d9f813ae6b42eadf580ec682a54a`. The checker reread its bytes. This establishes recorded-input/artifact concordance, not independent compiler provenance or completeness of undeclared/untracked dependencies.
|
|
|
|
Final full raw SHA256: `7cd8ac2de3b641e0421f48196be255c6b79706049e814e9f62ca2936794f8f71`. Final format raw SHA256: `c1c6b12861c8117ce6ffc6d6742ebf92618aafb10160ec6f99d370ca72ebab2d`. Exact per-gate raw/receipt hashes are in the readback JSON.
|
|
|
|
## Failure bodies: 15 exact plus one documented delta
|
|
|
|
After replacing only numeric thread PIDs, 15 inherited failure bodies are byte-identical. The remaining named block is `language::nesting_tests::checked_in_example_supports_cross_file_reparent_without_changing_allocations`. It lacks exactly one predecessor LF-terminated line: ``note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace``. Its panic site, error payload, diagnostic location and all remaining whitespace/bytes are unchanged. The amended checker requires this precise changed name, exactly one removed line, and the exact old/new normalized block hashes; it does not normalize other framework text.
|
|
|
|
Old normalized block SHA256: `cd6ec346dde1f43b82ca244a8b56754d209cf8108693caed8bb00a1a8ef32e1e`. New normalized block SHA256: `d074bcb3e4edbd2622b3cf1a0ed4cac1aa752f9761cb5ffc526540aed4fb1c04`. Initial6b0ef99 has this same delta and the other 15 exact blocks. A successful caught refund panic consuming Rust's process-once hint is plausible but remains an inference, not a traced cause. Do not summarize this evidence as 16 byte-identical failure bodies.
|
|
|
|
## Preserved initial candidate and stack limits
|
|
|
|
Initial `6b0ef99eca6f82a93e2cfc9eb16185c4f12403a7` passed the same six focused tests and completed full 259/16 and listing 275. Its Clippy exit 101 is preserved: `clippy::unnecessary_unwrap` at `language.rs:1227:25`, from checking `is_some()` then `expect()`. Initial Clippy raw SHA256 is `c4e70ff7795cc37041e5a4ae1b4028ec63530bc9917a00ef0dc25389b34d0564`; initial full raw SHA256 is `80f3ada1357adcbe213a43c60ae0aea62e8f5aaa900c3177c8c01ea96b341c17`. The successor diff is only the direct `if let Some(mixed)` rewrite and implementation-note update. Test bytes remain SHA256 `7b484b7356c48944ce3d1b1fc0e057b79bcfc643a2387978288de4305b381f20`. No lint allowance or test expectation changed.
|
|
|
|
Reviewed recorded initial/final stack and active-frame observations: selected reservations match between6b and97. Relative to8d, parse_impl increases512 bytes, Parser::unit208, accepted96, indexed64, worker256. Preserved binding2520 plus prefix11720 gives14240 bytes of those two selected reservations; continuation13416 is an alternative phase and must not be added to prefix. These are compiler prologue observations, not total runtime stack/headroom or arbitrary-input guarantees. Prior extra process rechecks and earlier parse/setup failures retain their original source/scope; none is relabeled as a run or repair of this checkpoint.
|
|
|
|
This accepts only the existing finite text-producer tests and the specified regression evidence. It does not establish all-heap accounting, a public production constructor, accepted rename output, GUI behavior, or runtime behavior outside these recorded gates.
|