# Identifying a CuePool build Help → About, startup logs, Status diagnostics and `/v1/health` use one source identity. Ordinary `cargo build` needs no environment setup. It records the workspace version, short and full Git commit, clean/modified state, nearest reachable `vX.Y.Z` tag and distance from it. Modified builds include a fingerprint of their diff and untracked files; this identifies local edits, but does not save a copy of them. Keep the checkout when diagnosing such a build. `cuepool --version` prints the same identity and ASIO build support without initializing the GUI, audio, logs or user profile. Windows release builds use the desktop subsystem so Start-menu, Explorer and project-file launches do not open a terminal. Version output preserves inherited pipes/files, and uses an existing parent console when needed; it never creates a console. Interactive `cmd.exe` does not wait for desktop applications, so use `start /wait` or PowerShell's `Start-Process -Wait -PassThru` when the exit status is required. Cargo checks Git every time it builds. This intentionally adds a small build-script cost: watching HEAD or the Git index alone misses new files, tag changes and some worktree operations. The generated data contains no timestamp and is only rewritten when its content changes. Source identity is stable for unchanged source. It does not promise byte-identical binaries across toolchains, platforms or build flags. `CUEPOOL_BUILD_ID` remains an optional packaging identifier. It is displayed in addition to detected source data and remains the value of the health API's existing `commit` field. Without an override, that field is the seven-character source SHA. The API additionally provides `build_identity`, `source_commit` (full SHA), `source_dirty`, `version_tag`, `commits_since_tag` and `tag_publication`. The existing `dirty` API field still describes the open show project, not source. Detached HEADs and Git worktrees work without special setup. Shallow clones show their commit but explicitly warn that tag/history information can be incomplete. Fetch full history and tags for complete comparisons. Archives without Git metadata report source identity unavailable; an explicit build identifier does not prove that the archive matches a release. ## Offline changes Help → Changes embeds maintained `CHANGELOG.md` entries and up to 60 first-parent commit descriptions at compile time. It states the comparison baseline, prefers a reachable confirmed published release, and otherwise labels its baseline as an unverified version tag. At an exact tag it compares against an earlier baseline. Local edits and incomplete history are explicitly disclosed. Commit history is grouped using the changed paths: application code/manifests, build maintenance/documentation, and other changes. Commit descriptions are attribution, not a guarantee of behavior. Empty history is never presented as proof that no application behavior changed. Maintained notes should explain operator impact concisely; the history retains PR numbers for further investigation. Confirmed publications are recorded by the publication workflow in annotated `published/vX.Y.Z` Git tags. These are receipts, not additional product versions. Their JSON annotation records the product tag, peeled commit, public URL and publication time. The product tag remains `vX.Y.Z`. Both must point to the same commit before CuePool describes the version tag as published. Fetch tags to obtain receipts; without a receipt publication is unverified, not assumed failed. The repository had no published releases when this was introduced. A tag from a failed packaging run is not a release. Publication is a snapshot at build time; a candidate built before publication cannot claim a future outcome. Rebuilding that commit after fetching its publication receipt can report it as published.