ref:ee6140d27b3bf410e37fb00d8156676dd7f333a7

docs(journal): record the CI runner findings and the broken-capture check

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BPNw4PCgkEfhyCjQT19wsb
SHA: ee6140d27b3bf410e37fb00d8156676dd7f333a7
Author: Cole Christensen <cole.christensen@gmail.com>
Date: 2026-09-13 12:43
Parents: e5117c6
1 files changed +9 -0
Type
JOURNAL.md +9 −0
@@ -676,3 +676,12 @@
- `.anvil.yml` plus `scripts/ci/anvil-test.sh`.
- The only online x86_64 runner is this WSL2 machine (the `anvil runner` service). Container steps get no
devices, so the KVM end-to-end step runs `image: bare` on the host.
- Runner behavior that shaped the config:
- It keeps the workspace between jobs, so builds are incremental (a full run takes 10 min instead of 32).
It also meant every job uploaded all of `ci-artifacts`, including ~900 Doxygen files, and hit the forge's
upload rate limit (429). Each job now uploads only its own files.
- An artifact `path` must be a file or a glob; a bare directory matches nothing.
- The host's distro QEMU is 8.2, so the reboot mode expects one display there.
- Run `4028172a3e` on `e5117c66`: all 6 jobs passed. dmabuf is reported as skipped (no `/dev/dri`).
- Broken capture check: with `convert_row` swapping red and blue, `anvil-test.sh e2e shm` exits 1
(`matched: false`, quadrants swapped). It exits 0 again once restored. The patch was never committed.