| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
ci: drop the weekly devcontainer rebuild The scheduled rebuild existed to re-sync the layer cache with a floating ubuntu:24.04 tag. With the base image pinned by digest the weekly run resolves the same base, hits cache on every layer and re-pushes an identical image. The push and pull_request path filters already refresh the cache whenever the devcontainer inputs change, which is the only time it now needs to happen. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Signed-off-by: Ryan Turner <ryan@turnrye.com> | 2 天前 | |
ci: support labelling prs and labels as 'In progress' to disable automation Signed-off-by: Ryan Turner <ryan@turnrye.com> | 1 个月前 | |
ci: check every string table against stringsTable_t directly Per turnrye's review on #497: the existing check only verified that translations stay in sync with EnglishStrings.h, which can't catch English itself drifting from stringsTable_t (e.g. a struct field added with no *Strings.h file updated to match). Adds a first pass that checks every *Strings.h file, English included, against the struct's own field list in ui_strings.h before the existing English-vs-translations pass runs. | 18 天前 | |
ci: keep push and pull_request runs in separate concurrency groups A push to a same-repo pull request branch starts both a push run and a pull_request run in the same concurrency group, and GitHub cancels the older pending run unconditionally. When the pull_request run loses, the PR build comment silently keeps offering the previous commit's firmware. Keying the group on the event name stops the two from cancelling each other while still cancelling superseded runs of the same event. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Signed-off-by: Ryan Turner <ryan@turnrye.com> | 1 天前 | |
ci: comment build artifact links on pull requests Reviewing a firmware change usually means flashing it, and finding the build for a pull request currently means digging through the Actions UI. The artifact links there also require a GitHub login, which many testers do not have. Post a comment on the pull request listing a download per radio, next to a link to the flashing guide. The comment is updated in place as new builds finish rather than added again, and is hidden when a build produces no firmware, so a failing build does not leave stale download links behind or add noise to a pull request that already shows a failing check. This runs as a separate workflow_run workflow because a pull_request run from a fork, which is how nearly all contributions arrive, gets a read-only token and cannot comment. It never checks out or runs the contributor's code. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Signed-off-by: Ryan Turner <ryan@turnrye.com> | 1 天前 | |
scripts: ci: use uv to install python dependencies Install the Python dependencies with `uv` instead of `pip`. This is a drop-in swap of the installer only: requirements.txt stays the source of truth, the scripts keep running against the same interpreter, and no build wiring changes. In the devcontainer this also lets uv supply the interpreter, which removes the deadsnakes PPA and the extra apt round-trip it needed. The environment is still a plain virtualenv at /opt/venv on PATH, and it still holds the same package set, so nothing downstream of it changes. python3-pip is dropped from the image, as nothing installs with pip any more. Adopting uv's project workflow -- pyproject.toml, a committed uv.lock and `uv run` -- is deliberately left out here so that this change stays small and easy to revert. That migration is worthwhile but it is a much larger diff, most of it the lockfile, and it makes uv a hard requirement of the build; it should be reviewed on its own rather than riding along with an installer swap. Signed-off-by: Ryan Turner <ryan@turnrye.com> Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> | 13 天前 |
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
| 2 天前 | ||
| 1 个月前 | ||
| 18 天前 | ||
| 1 天前 | ||
| 1 天前 | ||
| 13 天前 |