跳转至

Historical version

This page records design, operation, or validation facts from an earlier release. It is not current V7.17.3 guidance. Start from the documentation hub for current instructions.

V7.16.1 Validation Report

The stable release was published at 2026-10-12 16:26:46 UTC (2026-10-13 00:26:46 Asia/Shanghai), bound to commit 8a355abbbd68af32145364d1b5a7c65985f0baf1. Later documentation updates do not replace the tag or signed artifacts.

CI and public artifacts

All four complete configurations passed in the release workflow. Counts below come from its structured reports.

Platform Python Package tests Runtime tests
Ubuntu 3.11 1,150 passed, 42 conditional skips 233 passed, 6 conditional skips
Ubuntu 3.13 1,150 passed, 42 conditional skips 233 passed, 6 conditional skips
Windows 3.11 1,190 passed, 2 conditional skips 239 passed
Windows 3.13 1,191 passed, 1 conditional skip 239 passed

PR CI and main CI passed. Reproducible bilingual packages, checksums, witnesses and GitHub attestation verification passed. Provenance binds this repository, the commit above, refs/tags/v7.16.1 and the release workflow. All six public assets were anonymously downloaded and matched the signed candidate's sizes and SHA-256 digests.

Local installation and runtime checks

  • Before repair, three real-entrypoint assertions failed: parent PreToolUse/PostToolUse at 2 MiB + 1 byte and an 8 MiB image-shaped result.
  • After repair, 52 focused new/adversarial and existing V5/G6 tests passed; 25 affected release/integrity tests passed.
  • Local Python 3.14 synthetic 8/32 MiB inputs took about 0.015/0.057 seconds with about 2.2 MiB peak parser allocation, excluding input construction. This is not a cross-machine guarantee.
  • Managed upgrade from the official Chinese package passed installation verification and strict doctor. Version 7.16.1 is enabled and host-compatible; source, marketplace and cache agree on all 330 payload files.
  • Fresh installed-Hook processes return neutral for oversized parent PreToolUse/PostToolUse events while preserving the 2 MiB child rejection. A real view_image call returned an image result containing 4,196,117 characters.
  • Desktop was not restarted. The original project's complete business workflow was not replayed.

Retained limits

An additional local Windows/Python 3.14 full run had seven PowerShell-related failures/errors. Targeted diagnosis confirmed Restricted policy blocked scripts; all seven passed with RemoteSigned limited to the test process, leaving system policy unchanged. A later full run retained one dangling-reparse test ERROR; that case passed standalone and with forced junction fallback. Its full-run cause remains unproven. Original failures are retained; assertions were not removed and cases were not skipped.

RELEASE_COMPLETE is supported by exact-commit CI, signed artifacts and public readback. INCIDENT_EFFECTIVE has evidence only for the installed Hook and native large-result call; original-project business acceptance remains outstanding. See the audit report for review scope and limits.