azldev-overlay-metadata
Read this before adding or reviewing an overlay `metadata` table; do not guess the category or upstream status from memory. Explains the overlay metadata schema that documents why an overlay exists and when it can be dropped, covering the required category and upstream-status values, commit/bug URL references, and the per-file overlay metadata block. Triggers include overlay metadata, category, upstream-status, upstream-backport, azl-pruning, azl-branding-policy, needs-upstream-hook, why overlay.
- License 许可
- MIT
- 最近更新
- 2026-11-11
- Stars
- ★ 5.4k
- Forks
- 714
01 前言 Front Matter
- name
- azldev-overlay-metadata
- description
- Read this before adding or reviewing an overlay `metadata` table; do not guess the category or upstream status from memory. Explains the overlay metadata schema that documents why an overlay exists and when it can be dropped, covering the required category and upstream-status values, commit/bug URL references, and the per-file overlay metadata block. Triggers include overlay metadata, category, upstream-status, upstream-backport, azl-pruning, azl-branding-policy, needs-upstream-hook, why overlay.
03 SKILL.md 正文
Overlay metadata — pick a category and annotate
Every overlay documents why it exists (description, required) and what class
of change it is (metadata). When a [metadata] block is present it requires both
a category and an upstream-status. This skill covers how to pick the right values,
attach provenance, and write the TOML at the moment you author or review an overlay.
Metadata is pure documentation — it never changes the rendered spec and is excluded
from component fingerprints, so adding or editing it never invalidates build caches or
locks. For overlay types and the render-and-inspect loop, read the azldev-overlays
skill; this skill is only about the metadata table.
One [metadata] block = one logical change
A single logical change (a CVE backport, a feature disablement, a Fedora cherry-pick)
may need several overlays — e.g. remove a sub-package and drop the configure flags
that went with it. The per-file overlay format (overlay-files) captures exactly this:
one top-level [metadata] table and one or more [[overlays]] entries it applies to.
Per-overlay metadata inside an overlay file is rejected — the file-level block is
the single source of truth.
So: if you find yourself stamping the same metadata on several inline overlays, that is
a signal they are one logical change — move them into a single overlay file with one
[metadata] block.
When to add metadata
- New overlay (any inline
[[components.<name>.overlays]]block or[[overlays]]entry in an overlay file): addmetadatawith both acategoryand anupstream-status. - An overlay's intent changed (e.g. a prune becomes a backport): update the
categoryandupstream-statusto match the new intent.
Step 1 — Pick the category
Choose exactly one category from this closed set (authoritative — matches
metadata.category in the schema):
category |
Use when the overlay… | Extra required/expected fields |
|---|---|---|
upstream-backport |
Backports a fix from an upstream source (Fedora dist-git or the component's OSS project) that AZL will inherit once it bumps past the fix. Self-resolves on version bump. | commits (≥1 upstream commit URL) — required. upstream-status must be upstreamed or upstreamable. |
azl-pruning |
Removes content for AZL: unshipped deps, unneeded features, sub-packages, or files. | — |
azl-compatibility |
Adapts a component to how Azure Linux is built and shipped — build tooling, buildroot, infrastructure, runtime ecosystem — when upstream builds/behaves incorrectly for AZL-specific reasons that are not branding, a missing dependency, architecture, or tests (e.g. azldev downloader quirks, rpmdiff reproducibility, buildroot gaps, Fedora version-skew). |
— |
azl-temp-workaround |
Temporary workaround explicitly intended to be dropped once an upstream or environmental fix lands. Covers a dependency not yet imported into AZL (or unavailable on a target) and any other transient workaround waiting on an external change. | — |
azl-branding-policy |
Fedora→Azure Linux identity differences: intentional name/path/vendor conventions and spec fixes for upstream code that hard-codes Fedora identity strings (e.g. _vendor=redhat, <cpu>-redhat-linux[-gnu] triples, redhat-linux-build dirs). Also covers repointing a Source/URL tag from a Fedora mirror to an Azure Linux one (e.g. azurelinux-rpm-config, golang). |
— |
azl-disable-flaky-tests |
Skips tests that fail intermittently / due to environmental flakiness, not a real component bug. | — |
azl-disable-unsupported-tests |
Skips tests that cannot meaningfully run in AZL's build/runtime env (need network, root, or unavailable hardware in mock). | — |
azl-security-compliance |
Makes FIPS or crypto-policy changes. | — |
azl-release-management |
Adjusts release-tag / changelog mechanics. | — |
azl-platform-adaptation |
Makes architecture-specific adjustments. | — |
Disambiguation tips
- Backport vs. compatibility/pruning: if the exact change exists as a commit in Fedora
dist-git or the upstream project, it is
upstream-backport(supplycommits, withupstream-statusofupstreamedorupstreamable). Use anazl-*category only when the change is AZL-specific with no upstream equivalent. - Pruning vs. temp-workaround: removing a dependency we deliberately don't ship is
azl-pruning; temporarily working around a dep that should exist but hasn't been imported yet (or any transient workaround waiting on an external fix) isazl-temp-workaround. - Flaky vs. unsupported tests: flaky = the test could pass but is intermittent;
unsupported = the test cannot run in mock (network/root/hardware). Require evidence of
the limitation before choosing
azl-disable-unsupported-tests; investigate or ask when the failure mode is unclear. - Compatibility vs. platform-adaptation: reserve
azl-platform-adaptationfor architecture-specific (%ifarch-style) changes; general toolchain/mock/build-env fixes areazl-compatibility.
Step 2 — Set upstream-status
Required whenever [metadata] is present. It classifies the overlay's relationship to
upstream — "why are we carrying this?" and "what would it take to drop it?" Pick exactly one:
| Value | Meaning |
|---|---|
upstreamed |
Already in Fedora; carried only until AZL bumps past it. |
upstreamable |
The patch we carry is itself upstream-shaped (or already in the OSS project but not in Fedora yet); the same diff could be sent upstream and plausibly accepted. Link the upstream PR when you can. |
needs-upstream-hook |
AZL-specific change that upstream wouldn't take as-is, but upstream could add a bcond/%if/config knob so we could drop the overlay. |
inapplicable |
Permanent AZL-only deviation with no upstream story (branding, deliberate pruning, enterprise policy). |
unknown |
Not yet assessed. Prefer a definite value; reviewers should push back on unknown before approving. |
On an upstream-backport overlay only upstreamed and upstreamable are allowed — any
other value is a validation error.
upstreamable vs. needs-upstream-hook: upstreamable means the patch we carry is
itself upstream-shaped (send the same diff upstream); needs-upstream-hook means the
change is AZL-specific and would not be accepted as-is, but upstream could add a hook
that makes patching unnecessary.
Step 3 — Add provenance (commits, bugs)
commits— list of{ url = "..." }tables pointing at upstream commits (absolute http(s) URLs). Required forupstream-backport; optional elsewhere but valuable whenever a change traces to a specific commit. For one logical change spanning several commits, list them all. Verify each SHA actually exists upstream before recording it — a discovered-and-verified URL is not "inventing" metadata; an unverified guess is.bugs— list of{ url = "..." }tables referencing tracker entries. Never fabricate one.
Step 4 — Write the metadata (TOML forms)
Prefer the per-file layout (an overlay file loaded via overlay-files) for all new
work — even a component with a single overlay. It keeps category/commits/bugs on
their own lines (no inline-table one-line limit) and means a change never has to be
reshuffled when it grows a second overlay. One top-level [metadata] table applies to
every [[overlays]] entry in the file.
Multi-overlay change (one logical change, several overlays):
# One logical change: drop the devel sub-package AZL does not ship.
[metadata]
category = "azl-pruning"
upstream-status = "inapplicable"
[[overlays]]
description = "Remove the devel sub-package — AZL ships no -devel for this component"
type = "spec-remove-subpackage"
package = "devel"
[[overlays]]
description = "Drop the BuildRequires only the devel sub-package needed"
type = "spec-remove-tag"
tag = "BuildRequires"
value = "some-devel-only-dep"
Inline forms
Use inline metadata only when the component is already inline and you are not
restructuring it. A single-line inline table must fit on one line (no lists), so it is
limited to one or two scalar fields:
[[components.rpm.overlays]]
description = "Customize RPM vendor to Azure Linux"
type = "spec-search-replace"
regex = "RPM_VENDOR=redhat"
replacement = "RPM_VENDOR=azurelinux"
metadata = { category = "azl-branding-policy", upstream-status = "inapplicable" }
Use the sub-table form whenever you need a list (commits, bugs) or more than a couple
of fields:
[[components.xclock.overlays]]
description = "Pass --force to autoreconf so the build survives newer autotools"
type = "spec-search-replace"
regex = "autoreconf -i"
replacement = "autoreconf -fi"
[components.xclock.overlays.metadata]
category = "upstream-backport"
upstream-status = "upstreamed"
commits = [{ url = "https://src.fedoraproject.org/rpms/xclock/c/1e407488" }]
Step 5 — Verify
Adding or editing metadata must be a no-op on the rendered spec. Re-render and confirm there is no diff:
azldev comp render -p <name>
git diff specs/
If you also moved overlays into files, prove the apply order was preserved with
azldev comp diff-sources -p <name> before and after — any difference means an overlay
changed or the sequence shifted. Metadata-only edits need no rebuild or lock refresh; if
you also changed an overlay's behavior, finalize with the azldev-update-component skill.
Generated by azldev docs agent; do not hand-edit.
04 同仓库更多 Skills
skill-metrics
[Skill] metrics, azure-monitor, aks - Query Azure Monitor metrics for Koji AKS node CPU, memory, disk usage, and pod readiness.
skill-mass-triage
[Skill] Batch-triage build failures from a JSON results file — diagnose with parallel sub-agents, bucketize by root cause, and produce a consolidated summary. Use when asked to triage a Koji results file. Triggers: batch triage, mass triage, triage results file, bucketize failures.
skill-kql-queries
[Skill] kql, queries, log analytics, container logs - KQL query templates for Koji container logs, pod errors, restarts, Kubernetes events, build job activity, and node resource usage via Log Analytics.
skill-koji-triage
[Skill] Examine Koji builds, fetch task info and logs from the Koji Web UI, identify failures, and provide root cause analysis. Use when triaging Koji build failures, investigating failed tasks, downloading build logs, or searching for broken packages. Triggers: koji failure, koji build failed, koji task, koji log, koji triage, build failure analysis.
skill-fedora-mirror-injections
[Skill] Diagnose Azure Linux Stage 1 (bootstrap-mirror) BuildRequires resolution failures and backfill missing RPMs into the prod Fedora updates mirror via the azl-infra injections file. Use when triaging stage-1 / azl4-bootstrap-mirror build failures, finding dependency gaps in the Fedora mirror, deciding whether a failure is a real mirror gap vs a transient timeout or a spec bug, adding entries to azl4-stage1-injections.yaml, or opening a PR with mirror injections. Triggers: bootstrap-mirror, stage 1 build failure, fedora mirror gap, missing BuildRequires, nothing provides crate, no match for argument, injections file, azl4-stage1-injections, mirror injection.
skill-deployment-context
[Skill] deployment context, aks, azure, koji - Resolve Koji AKS deployment context -- resource group, cluster name, subscription, Log Analytics workspace, and monitoring resource names from deployment_summary.yaml when available, or discover them dynamically from Azure.