[AIwork4me]

What & why

Documentation consistency audit (2026-08-27, independent reviewers over every user-facing doc) found the docs one step behind the 2026-08-26 upstream events, plus a CI collection break. This PR brings every document back in line with the latest verified facts.

• PR#260 retarget status everywhere: upstream main absorbed the 10-call-site fix (2f42002); PR#260 retargeted to main as a 1-line commit (5f4aca0e), approved by yl-1993 2026-08-26. Updated: docs/porting.md §6 (EN+CN — including the closing advice, which now correctly conditions patch removal on the pinned checkout advancing past 2f42002, not on "upstream merging"), patches/README 0001 status line, the upstream tracking doc, and a CHANGELOG entry (the retarget commit itself had none).
• Wave-1 visibility: README/README_CN script inventory now lists scripts/rocm_check.py and scripts/gpu_monitor.py; wave-1 design spec gets a dated addendum noting 8060S's completed alignment wave-2; archived pytorch#194447 copy gains its Resolution (2026-08-26) banner; README_CN gains the CI badge (EN/CN parity).
• CI repair: tests/validation/test_pytorch_nightly_bug{1,2,3}.py import torch at module level → pytest tests/ failed collection on torch-less runners (main's CI red on the last 3 push runs). Now pytest.importorskip("torch") — they run on GPU hosts, skip on CI.

Evidence

• Audit verified against live upstreams on 2026-08-27 (PR#260 open/retargeted/approved; #194447 closed; #194498 open; rocm-systems#10685 open).
• Local gates: pytest tests/ -q → 1 failed (pre-existing environmental test_fullstack_mode_logic, documented), 10 passed, 3 skipped (the guarded GPU repros); shellcheck clean. Docs-only diff otherwise.

Gates

• python3 -m pytest tests/ -q (see note above — the environmental failure reproduces on pre-branch main)
• shellcheck scripts/*.sh scripts/lib/common.sh docs/hardware/strix-halo/hsa_fix.sh