Struct ContextOptions 

pub struct ContextOptions {

    pub snapshot: bool,
    pub nightly: bool,
    pub dry_run: bool,
    pub quiet: bool,
    pub verbose: bool,
    pub debug: bool,
    pub skip_stages: Vec<String>,
    pub selected_crates: Vec<String>,
    pub token: Option<String>,
    pub parallelism: usize,
    pub single_target: Option<String>,
    pub release_notes_path: Option<PathBuf>,
    pub fail_fast: bool,
    pub partial_target: Option<PartialTarget>,
    pub merge: bool,
    pub publish_only: bool,
    pub project_root: Option<PathBuf>,
    pub strict: bool,
    pub resume_release: bool,
    pub replace_existing_artifacts: bool,
    pub skip_post_publish_poll: bool,
    pub gate_submitter: Option<bool>,
    pub rollback_mode: Option<RollbackMode>,
    pub simulate_failure_publishers: Vec<String>,
    pub rollback_only: bool,
    pub allow_rerun: bool,
    pub show_skipped: bool,
    pub from_run: Option<String>,
    pub runtime_nondeterministic_allowlist: Vec<(String, String)>,
    pub summary_json_path: Option<PathBuf>,
    pub allow_ai_failure: bool,
    pub changelog_from: Option<String>,
    pub changelog_full_history: bool,
    pub changelog_to: Option<String>,
    pub changelog_preview: bool,
}

Fields

snapshot: bool

nightly: bool

dry_run: bool

quiet: bool

verbose: bool

debug: bool

skip_stages: Vec<String>

selected_crates: Vec<String>

token: Option<String>

parallelism: usize

Maximum number of parallel build jobs (minimum 1).

single_target: Option<String>

When set, build only for this single host target triple.

release_notes_path: Option<PathBuf>

Path to a custom release notes file (overrides changelog).

fail_fast: bool

When true, abort immediately on first error during publishing.

partial_target: Option<PartialTarget>

Partial build target for split/merge mode. When set, the build stage filters targets to only those matching this partial target.

merge: bool

When true, running with --merge flag (merging artifacts from split builds).

publish_only: bool

--publish-only: load artifacts from a preserved dist (written by anodize check determinism --preserve-dist=...) and run only the sign + publish pipeline. The CLI dispatcher uses this flag in setup_env to defer the GitHub-token check to publish_only::preflight_credentials, which owns the combined token + sign-key check and bails fail-closed on missing values. Without this gate, setup_env’s token check would fire FIRST and pre-empt publish-only’s own preflight (which validates BOTH token AND sign key in one shot).

project_root: Option<PathBuf>

Explicit project root directory. When set, stages use this instead of discovering the repo root via git rev-parse --show-toplevel.

strict: bool

Strict mode: configured features that would silently skip become errors.

resume_release: bool

--resume-release: opt-in to continue into a release left over from a prior failed attempt. Bypasses the leftover-assets pre-check that bails when an existing release already has assets and replace_existing_artifacts is false.

replace_existing_artifacts: bool

--replace-existing: CLI override that forces release.replace_existing_artifacts: true regardless of config. The release stage ORs this with the config value.

skip_post_publish_poll: bool

--no-post-publish-poll: skip post-publish polling for the chocolatey moderation queue and the winget PR validation pipeline. When true, the polling runner emits PostPublishStatus::NotPolled (pending immediately) for every publisher rather than waiting on a terminal state. Lets CI users with no patience for long-running waits opt out without scattering post_publish_poll.enabled: false across every publisher block.

gate_submitter: Option<bool>

Whether the publisher dispatcher gates irreversible Submitter publishers (chocolatey, winget, AUR-source, krew, snapcraft) on the success of every required Assets/Manager publisher that ran before them. None defaults to Some(true) (gate on). The CLI flag --no-gate-submitter flips this to Some(false). See stage-publish::dispatch::DispatchOptions::gate_submitter for the gating mechanics.

rollback_mode: Option<RollbackMode>

--rollback=<none|best-effort>: post-publish rollback policy. None means “resolve from preflight state at dispatch time” (best-effort when preflight ran clean, none otherwise with a warn). Consumed by the rollback-dispatch task.

simulate_failure_publishers: Vec<String>

--simulate-failure=<publisher> (repeatable, hidden, env-gated behind ANODIZE_TEST_HARNESS=1): names of publishers whose run() should be skipped and a synthetic Failed("simulated failure: <name>") recorded in the report instead. Lets the failure-mode test harness exercise gate / rollback / report paths deterministically without monkey-patching production publisher code.

rollback_only: bool

--rollback-only: skip publish; re-attempt rollback from a prior run report. Requires from_run to identify which prior run’s report.json to load. The actual replay logic lands in a follow-up task; this field is plumbed so the flag is visible in --help today.

allow_rerun: bool

--allow-rerun: force PublishStage::run to proceed even when a prior report.json exists for the current run_id. The default (false) refuses re-runs to prevent PR-based publishers (homebrew / scoop / nix / krew / MCP) from duplicating their pull requests against the same tag.

Operators recovering from a partial failure should prefer --rollback-only --from-run=<id> (which has its own idempotency guard via dist/run-<id>/rollback.json). The rerun flag is an escape hatch for advanced cases where the operator has manually verified no duplicate-publish risk exists.

Audit ref: 2026-05-15 release-resilience-review finding I4.

show_skipped: bool

--show-skipped: surface the per-crate “no <publisher> config block” skip lines at default verbosity. In workspace mode every PR-based publisher (homebrew / nix / scoop / aur / winget / krew / chocolatey) visits every selected crate and skips the ones whose config lacks its block; at default verbosity those no-op skips are routed to debug (invisible unless --debug) so they do not bury the real output. Setting this flag forces them back to status — the diagnostic escape hatch for “why didn’t publisher X run for crate Y?”.

from_run: Option<String>

--from-run=<id>: prior run id whose report.json to load when running in --rollback-only mode. clap enforces the requires = "rollback_only" relationship at parse time.

runtime_nondeterministic_allowlist: Vec<(String, String)>

--allow-nondeterministic <name>=<reason> (repeatable): runtime non-determinism opt-outs for specific artifacts. The determinism stage suppresses its non-determinism error for any matching artifact name, recording the supplied reason in the report. Mutually exclusive with --strict at the clap layer.

summary_json_path: Option<PathBuf>

--summary-json=<path>: when set, the per-publisher run summary is written to this path. Consumed by the run-summary task.

allow_ai_failure: bool

--allow-ai-failure: when true, a failure inside the changelog.ai enhancement step (transport, non-2xx, parse) is logged as a warning and the pre-AI release notes are kept verbatim. Default false (fail-closed) follows the conventional “any hook failure aborts” pattern: a silent fall-back to the raw notes ships the wrong body without the operator noticing.

changelog_from: Option<String>

changelog --from <ref>: explicit lower bound (range start) for changelog commit collection. When set, the changelog stage uses this ref as the previous tag instead of auto-discovering the latest matching tag. A dedicated option (rather than the always-auto-populated PreviousTag template var) so a full release run’s per-crate auto-discovery is never overridden — only an explicit --from is.

changelog_full_history: bool

changelog .. / changelog ..<ref>: an explicit empty lower bound, meaning “from the beginning of history” with no auto-discovered previous tag. When true, the changelog stage skips tag auto-discovery entirely so the range covers all reachable commits up to the upper bound — distinguishing the explicit empty-from form from an omitted range (which still resolves to the last release tag).

changelog_to: Option<String>

changelog <from>..<to> / changelog <tag>: an explicit UPPER bound (range end) for changelog commit collection. When set, the changelog stage walks <from>..<to> instead of <from>..HEAD, so commits AFTER <to> are excluded. A dedicated option (rather than the always-populated Tag template var) so the pending / snapshot window — where Tag resolves to the latest EXISTING tag yet the range must still run to HEAD — is never silently bounded to that tag. None keeps the upper bound at HEAD (the pending window since the last release).

changelog_preview: bool

Marks the run as the standalone changelog --format release-notes LOCAL preview, NOT the release/tag pipeline. The standalone command is an inspection tool: it must render the pending window from local git with no release-time preconditions, so this flag relaxes three guards that are correct for a real release but wrong for a preview:

• the tag-must-point-at-HEAD + dirty-tree bails in resolve_git_context (a preview must not require a checkout or a clean tree),
• the snapshot-skip config gate in the changelog stage (a preview must render without changelog.snapshot: true),
• the use: github-native branch (a preview renders from local git instead of requiring a token / emitting empty bodies).

ONLY the standalone changelog command sets this; the release/tag pipelines leave it false so their guards stay fully intact.

Trait Implementations

impl Default for ContextOptions

fn default() -> Self

Returns the “default value” for a type.