Documentation-system implementation and formal acceptance¶
All 58 tasks in this documentation-system iteration completed independent acceptance against their original exit conditions. Actual records cover implementation, tutorials, diagrams, automated gates, locked tools, formal builds, public deployment and remote recovery. The 58-task source ledger is the sole authority for task values; the readable checklist and original fields by task are generated from it.
The delivery results index records exact commits, archive digests, download locations, evidence members and per-task acceptance grounds. Here, passed means the corresponding documentation exit condition was accepted. It does not change the source, configuration, tools or physical status of original test records, or mean that this documentation commit reran all historical firmware tests.
Formal delivery and actual validation¶
The primary documentation qualification is bound to clean source 3835e13cd7da34bd75307746859d4514745ae2ff, version docs-20261014-3835e13cd7da. Later ledger and delivery-index commits record those actual results and have their own build manifests; their HEAD must not be inserted into earlier execution records.
Evidence scope |
Actual result |
|---|---|
Strict build |
Doxygen, gettext and English/Chinese HTML completed; Sphinx reported zero warnings |
Content and translation |
99 maintained documents and five content templates; 273 translations in 22 critical-page and navigation catalogs independently reviewed |
Architecture diagrams |
17 DOT/SVG pairs, including 15 required diagram classes; source, arrows, states, explanations and local fonts checked |
Browser and print |
162 checks, 15 SVGs and 22 A4 PDFs; 32 code blocks and 800 code lines complete; 18 pages in six PDFs actually inspected |
Reader tasks |
85 top-level commands and 28 SDK internal commands executed at clean |
Commit gates |
Normal hooks passed; 177 GoogleTest/GoogleMock executables with 3,252 cases; 1,555 cases across four tool domains; another 169 cases executed for documentation |
Public-site comparison |
All 1,895 real HTTP requests returned 200; 1,600 payload files, 290 legacy-URL root entries, stable and the actual downloaded archive matched byte for byte |
Remote recovery |
Nine assets actually downloaded from fixed Git commits; 1,600 files independently restored; 75 tool archives and two configurations matched; another 162 browser checks and 22 new PDF checks executed after recovery |
Per-task closure |
All 58 original exit conditions, 230 source migrations, 287 legacy URLs and original qualification boundaries independently checked; IDs and exit conditions preserved verbatim |
The source used for reader-task execution remains 7aadd347. Actual Git comparison proved that only print CSS changed between it and the primary documentation-qualification source; relevant code, configuration and tool subtrees are identical. Printing was separately accepted at 3835e13c. Normal commit-hook records retain each staged scope and original execution identity; they are not relabeled as new firmware qualification for a later clean HEAD. Newly generated PDFs after recovery passed mechanical and complete-code checks. The original 18-page visual results were reused through per-SHA equality of all 1,600 rendering inputs, without claiming those pages were inspected again.
Online reading, offline packages and long-term evidence¶
The online entry selects the current published version. Immutable versions/<release_id>/ retains original bytes; old bookmarks at the site root show accurate historical or current-contract explanations. Legacy HAL/OSAL APIs are not silently mapped to current interfaces.
The documentation release notes provide fixed-commit access to offline HTML, the complete documentation-tool cache and execution evidence. Actual long-term storage uses independent Git asset commits. The Release-attachment API returned 401; Git files were not described as uploaded attachments. The evidence index records actual upload, redownload and member verification results, while retaining original failed attempts.
Check bytes and SHA-256 before using an archive. Fetch multipart archives in the recorded order; each part is at most 32 MiB. Verify each part before reconstruction, then verify the whole archive and every inventory member. Receipts retain original commands, exit codes and original host paths; portable locators map those paths to relative members. Separately retained tool archives and documentation packages are not duplicated in the evidence package. An explicit exclusion table records each file’s digest, immutable download location and archive member, preserving original execution records completely.
Retained tags and later commits keep original versions and asset commits reachable. New versions add identities without replacing same-name archives. Maintainers keep independent copies according to the release notes. Digests detect byte differences; they are not digital signatures. See the documentation delivery contract for exact rejection rules.
Applicable scope and platform qualification¶
The documentation-tool closure was actually installed and verified on Linux x86_64/Python 3.12, using Chromium 151 on Linux. Local HTTP, browser offline reading after server shutdown, search, language switching and local diagrams were actually exercised. Existing enterprise policy blocked file://; the policy was unchanged and that path is not claimed as accepted. Noncritical translations display original/missing/stale/reviewed per page; bilingual entry points do not mean every historical page has been translated.
Platform firmware qualification still references cc6a065c and the original exact SoC indexes; see the software-verification entry. MCU HIL remains not_executed. Documentation builds, publication, recovery and software-model checks add no chip, board, product-safety or physical qualification.