feat(contracts): migrate iac.provider to strict gRPC proto contracts + CI validation by Copilot · Pull Request #41 · GoCodeAlone/workflow-plugin-digitalocean

Copilot · 2026-05-01T06:12:27Z

The plugin advertised iac.provider with no strict contract descriptor, failing strict plugin validation with missing_module_contract_descriptor.

Changes

Added proto/digitalocean.proto and generated Go bindings for IacProviderConfig; documented pinned protoc v25.3 / protoc-gen-go v1.34.2 versions in proto/generate.go for reproducible regeneration.
Implemented sdk.TypedModuleProvider and sdk.ContractProvider for the DigitalOcean IaC provider while preserving the legacy sdk.ModuleProvider fallback (compile-time assertion included).
Added plugin.contracts.json and included it in release archives via .goreleaser.yaml.
Added strict contract CI validation that validates plugin.json both as checked-in and after simulating the GoReleaser sed rewrites (mirrors .goreleaser.yaml before-hooks), plus release-time validation after GoReleaser rewrites plugin.json; both derive wfctl from the Workflow SDK version pinned in go.mod.
Kept checked-in plugin.json download URLs pointing to the latest published v0.8.4 hyphenated archives; GoReleaser before-hooks rewrite the version and URLs atomically at release time.
Moved the gRPC end-to-end strict-contracts test (TestPlugin_GRPCStrictContractsEndToEnd) to internal/plugin_integration_test.go with //go:build integration so go test ./... stays hermetic; integration tests run with -tags integration.
Added TestPluginDownloadsMatchGoReleaserArchives to cross-check download URL format and verify GoReleaser archive files list includes plugin.contracts.json.
Added TestPlugin_StaticContractManifestMatchesRuntimeRegistry to keep plugin.contracts.json in sync with ContractRegistry().
Merged current main to clear the PR conflict and retain the latest DigitalOcean provider updates.

Validation

GOWORK=off go test ./...
GOWORK=off go test -race ./...
GOWORK=off go vet ./...
GOWORK=off go vet -tags integration ./internal/
WFCTL_VERSION="$(go list -m github.com/GoCodeAlone/workflow | awk '{print $2}')" && GOWORK=off go run "github.com/GoCodeAlone/workflow/cmd/wfctl@${WFCTL_VERSION}" plugin validate --file plugin.json --strict-contracts

Strict validation now reports:

OK  workflow-plugin-digitalocean v0.8.4 (plugin.json)

Copilot

Copilot wasn't able to review any files in this pull request.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

intel352 · 2026-05-01T06:31:18Z

@copilot This PR currently has 0 changed files, so the assigned task is not implemented yet. Please continue the original issue task using Claude Sonnet 4.6: make the required Workflow strict-contract or downstream dependency updates, push actual code/config/doc changes to this PR branch, run the requested validation, and update the PR body with the validation results. Do not leave this as an empty PR.

intel352 · 2026-05-01T06:39:34Z

@copilot Add this as part of the PR scope: include or update a GitHub Actions PR check that validates strict Workflow plugin contracts. The check should fail when plugin.json is missing and should run the current Workflow CLI strict validation, for example:

name: wfctl strict contracts
on:
  pull_request:
  push:
    branches: [main, master]

jobs:
  wfctl-strict-contracts:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-go@v5
        with:
          go-version-file: go.mod
      - name: Validate strict plugin contracts
        run: go run github.com/GoCodeAlone/workflow/cmd/wfctl@v0.20.1 plugin validate --file plugin.json --strict-contracts

Adjust branch filters and Go setup to match this repository. If the repo uses an existing CI workflow, add this as a job there instead of creating duplicate workflow clutter. Include the check result in the PR validation notes.

Agent-Logs-Url: https://github.com/GoCodeAlone/workflow-plugin-digitalocean/sessions/7ff39de7-6d6d-4104-be5b-5685d34865c5 Co-authored-by: intel352 <77607+intel352@users.noreply.github.com>

- Add proto/digitalocean.proto with IacProviderConfig message - Generate proto/digitalocean.pb.go via protoc - Implement sdk.TypedModuleProvider on doPlugin (strict proto config path) - Implement sdk.ContractProvider on doPlugin (GetContractRegistry RPC) - Add plugin.contracts.json with strict module descriptor for iac.provider - Fix plugin.json: tier community, downloads, minEngineVersion 0.20.1 - Upgrade workflow SDK v0.19.0 → v0.20.1 - Add tests: ContractRegistry, TypedModuleTypes, CreateTypedModule variants Agent-Logs-Url: https://github.com/GoCodeAlone/workflow-plugin-digitalocean/sessions/7ff39de7-6d6d-4104-be5b-5685d34865c5 Co-authored-by: intel352 <77607+intel352@users.noreply.github.com>

intel352 · 2026-05-01T11:55:44Z

@copilot please try again using Claude Sonnet 4.6.

Continue the original task for this PR. Make sure the PR is not left empty, add or update strict gRPC contract descriptors as applicable, add a PR check that runs wfctl strict validation, run the repository tests, and keep this PR ready for review with Copilot assigned as reviewer.

Agent-Logs-Url: https://github.com/GoCodeAlone/workflow-plugin-digitalocean/sessions/f5a75b7b-047c-4e2c-954d-70bb922b6e07 Co-authored-by: intel352 <77607+intel352@users.noreply.github.com>

intel352 · 2026-05-02T20:26:23Z

Addressed the latest review in 08d0d9e: PR description now matches v0.8.4, module type duplication is centralized in code and cross-checked against static manifests, generated proto now has a go:generate command, release/download matrix is tested against GoReleaser config, and the gRPC strict path test now proves legacy nil-config creation would fail while the adapter-created strict module succeeds. Re-ran go test ./..., go test -race ./..., go vet ./..., and strict wfctl validation.

Copilot

Pull request overview

Copilot reviewed 10 out of 10 changed files in this pull request and generated 3 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

intel352 · 2026-05-02T23:27:17Z

Addressed the latest package/release review in f97dec3: go:generate now regenerates from repo root using proto/digitalocean.proto so the descriptor source path stays stable; the gRPC end-to-end test stages plugin.contracts.json and runs strict wfctl validation against the staged plugin package; and the GoReleaser validation hook exports GOPRIVATE and GOWORK=off before invoking wfctl. Re-ran go test ./..., go test -race ./..., go vet ./..., focused staged-package test, and strict wfctl validation.

Copilot

Pull request overview

Copilot reviewed 10 out of 10 changed files in this pull request and generated 6 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

intel352 · 2026-05-02T23:39:31Z

Addressed the latest release-archive review in 93673af: plugin.json now points at the next release tag v0.8.5 so installs will not keep using pre-contract v0.8.4 archives; tests now check GoReleaser archive files include plugin.contracts.json and build a representative tar.gz containing it; the gRPC test disables the legacy module path to prove strict typed dispatch; and TestPlugin_Manifest now catches runtime/plugin.json metadata drift. Re-ran go test ./..., go test -race ./..., go vet ./..., focused package tests, and strict wfctl validation.

Copilot

Pull request overview

Copilot reviewed 10 out of 10 changed files in this pull request and generated 5 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

+  "version": "0.8.5",
  "description": "DigitalOcean IaC provider: App Platform, DOKS, databases, Redis cache, load balancers, VPC, firewall, DNS, Spaces, DOCR, certificates, Droplets, IAM (declared), and API gateway",


+      - name: Validate strict plugin contracts
+        run: |
+          WFCTL_VERSION="$(go list -m github.com/GoCodeAlone/workflow | awk '{print $2}')"
+          go run "github.com/GoCodeAlone/workflow/cmd/wfctl@${WFCTL_VERSION}" plugin validate --file plugin.json --strict-contracts


+	if os.Getenv("WORKFLOW_PLUGIN_DIGITALOCEAN_DISABLE_LEGACY_MODULE") == "1" {
+		return nil, fmt.Errorf("digitalocean plugin: legacy module path disabled")


+		t.Fatalf("write temp plugin.contracts.json: %v", err)
+	}
+
+	validate := exec.Command("go", "run", "github.com/GoCodeAlone/workflow/cmd/wfctl@"+workflowModuleVersion(t, repoRoot), "plugin", "validate", "--file", filepath.Join(pluginDir, "plugin.json"), "--strict-contracts")


@@ -0,0 +1,3 @@
+package proto
+
+//go:generate sh -c "cd .. && protoc --go_out=. --go_opt=paths=source_relative proto/digitalocean.proto"


intel352 · 2026-05-03T01:20:37Z

@copilot address all unresolved review comments

- plugin.json: revert version/URLs to v0.8.4 (latest published); goreleaser rewrites to release tag at release time, avoiding 404 during pre-release window - .github/workflows/ci.yml: add post-rewrite simulation step that applies the same sed rewrites as goreleaser before-hooks and validates the result, so a broken template change is caught before release - internal/plugin.go: remove WORKFLOW_PLUGIN_DIGITALOCEAN_DISABLE_LEGACY_MODULE env kill-switch from production CreateModule; backward-compat hosts are not affected by test internals - internal/plugin_test.go + plugin_integration_test.go: move TestPlugin_GRPCStrictContractsEndToEnd and workflowModuleVersion to a //go:build integration file so go test ./... stays hermetic; remove t.Setenv kill-switch usage now that env var is gone from production code - proto/generate.go: document pinned protoc v25.3 / protoc-gen-go v1.34.2 versions used to generate the checked-in .pb.go Agent-Logs-Url: https://github.com/GoCodeAlone/workflow-plugin-digitalocean/sessions/55049097-392b-4939-aa13-b98c931fdb0f Co-authored-by: intel352 <77607+intel352@users.noreply.github.com>

Agent-Logs-Url: https://github.com/GoCodeAlone/workflow-plugin-digitalocean/sessions/55049097-392b-4939-aa13-b98c931fdb0f Co-authored-by: intel352 <77607+intel352@users.noreply.github.com>

Copilot · 2026-05-03T01:35:57Z

@copilot address all unresolved review comments

All 5 unresolved review threads addressed in commits 17c23a5 + 945ba40:

plugin.json:3-4 — pre-release 404s: Reverted version/download URLs from v0.8.5 to v0.8.4 (latest published). GoReleaser before-hooks rewrite to the actual release tag atomically at release time, so the checked-in manifest always points to existing artifacts.
.github/workflows/ci.yml — CI doesn't validate post-rewrite manifest: Added a second wfctl validation pass that applies the same sed rewrites as goreleaser's before-hooks to a temp copy of plugin.json, then validates that rewritten form. Uses portable sed -i.bak + cleanup.
internal/plugin.go:65-66 — env kill-switch in production code: Removed WORKFLOW_PLUGIN_DIGITALOCEAN_DISABLE_LEGACY_MODULE check from CreateModule. The env var was test-only logic that could break backward-compat hosts at runtime.
internal/plugin_test.go:171 — non-hermetic go run ...@version in test suite: Moved TestPlugin_GRPCStrictContractsEndToEnd and workflowModuleVersion to internal/plugin_integration_test.go with //go:build integration. go test ./... is now hermetic; integration tests run only with -tags integration.
proto/generate.go:3 — protoc versions not pinned: Added package doc documenting protoc v25.3 / protoc-gen-go v1.34.2 (exact versions used to generate the checked-in .pb.go) and instructions for regeneration.

…side plugin.json The Validate strict plugin contracts step (added in PR #41) copies only plugin.json to a tempfile when simulating GoReleaser's before-hook version rewrite. But wfctl plugin validate --strict-contracts looks for plugin.contracts.json CO-LOCATED with plugin.json (same directory). The rewrite-validation therefore always fails with "missing_module_contract_descriptor: module type \"iac.provider\" has no strict contract descriptor" — even when plugin.contracts.json IS valid in the repo root. This was masked on main because main has a single configurable contract file and any change rebuilds in-place; surfaced now by PR #55 which introduces a new resource type and exercises the second validation path. Fix: use a temp DIR and copy both plugin.json + plugin.contracts.json into it so the rewritten plugin.json's neighbour is intact. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…+ new infra.volume driver (#55) * feat(drivers): add boolFromConfig and strSliceFromConfig helpers Both helpers are structpb-aware: boolFromConfig falls through to the default for non-bool values, and strSliceFromConfig accepts the typed []string shape Go-native callers emit AND the []any shape that survives a YAML/JSON → structpb round-trip. Empty strings drop silently. Used by the upcoming droplet/volume drivers; placed in util.go alongside the existing strFromConfig/intFromConfig helpers so all drivers can reach them without import cycles. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat(droplet): support user_data / vpc_uuid / ssh_keys / volumes / tags / bools Self-hosted services (e.g. Apache-AGE Postgres on a Droplet, since DO Managed Postgres has no AGE extension) need more than the size/image/ region the DropletDriver previously honoured. New optional config keys, all additive and defaulted-empty (no behaviour change for existing callers): - user_data string — cloud-init payload - vpc_uuid string — VPC the Droplet joins - ssh_keys []string|[]int — fingerprints OR numeric IDs; element type detected at runtime (mixed lists OK; structpb floats must be whole numbers) - tags []string - enable_backups bool — godo Backups field - monitoring bool - ipv6 bool - volumes []string of — names resolved to IDs at create volume names time via Storage.ListVolumes(name=, region=). Region-bound: a name not present in the droplet's region returns "droplet volumes: volume %q not found". Outputs gain `private_ip` (godo droplet.PrivateIPv4()) so downstream services in the same VPC can be wired without a second Read. DropletDriver now holds an optional Storage client (set by NewDropletDriver from c.Storage; tests inject via the variadic NewDropletDriverWithClient param). When `volumes` is non-empty but the storage client is nil, Create returns an explicit error rather than silently dropping the attachment. mockDropletClient now captures the godo.DropletCreateRequest so tests can assert on populated struct fields. Adds focused tests for each new key, plus the SSH-keys fractional-float rejection path that prevents silent ID truncation. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat(drivers): add infra.volume driver (DO Block Storage) New ResourceDriver covering CRUD + Diff + HealthCheck for DigitalOcean Block Storage volumes. Backed by godo.StorageService; in-place size growth uses godo.StorageActionsService.Resize. ProviderIDFormat is UUID (DO volume IDs are 36-character UUIDs). Config keys: - name from spec.Name - region defaults to provider region - size_gb required, > 0 - filesystem_type optional ("ext4" / "xfs"); empty = raw block device - description optional - tags optional []string Outputs: id (UUID), name, region, size_gb, filesystem_type. Status is a stable "available" string — godo.Volume exposes no Status field, so a successful Read on a real volume is the strongest health signal we have. Update semantics: - size growth → in-place StorageActions.Resize, then re-read for state - size shrink → explicit error (DO has no shrink API; replace required) - region or filesystem_type change → Diff sets ForceNew so plan classifies as replace, not update Tests cover Create / Read / Update (resize, no-op, shrink-rejected, empty-ProviderID guard) / Delete / Diff (nil current, grow=update-only, shrink=replace, region=replace, filesystem_type=replace, no-changes) / HealthCheck (success, read-error, empty-ProviderID) / ProviderIDFormat / Scale-not-supported. The volume mock pool supports filter-by-name so the same struct exercises both VolumeDriver tests and the droplet volumes-by-name path. providerid_format_test.go adds the volume entry to the manually maintained registry — without it, a future drift in ProviderIDFormat would not be caught. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat(provider): register infra.volume driver and capability Wire VolumeDriver into Initialize() alongside the other infra drivers and declare its IaCCapabilityDeclaration with the noScale operation set (volumes have no scale concept; size_gb is dimensional). Add "infra.volume" to the sizing map as a noopSizing entry so ResolveSizing returns "n/a" for every abstract tier — the operator must set size_gb explicitly. Provider Capabilities test gains the matching required entry so removal of the driver registration would be caught. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * docs: declare infra.volume in plugin.json and CHANGELOG plugin.json: add infra.volume to the iacProvider.resourceTypes array and update the descriptor to mention Block Storage volumes. CHANGELOG: document the new infra.volume driver and the extended Droplet config keys (user_data / vpc_uuid / ssh_keys / volumes / tags / enable_backups / monitoring / ipv6) plus the new private_ip Droplet output. No README exists in this repo, so the CHANGELOG is the canonical surface for these notes. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(droplet): strict validation for volumes config entries (Copilot finding #1) strSliceFromConfig silently drops non-string and empty entries, which is dangerous for volume attachments — a typo or wrong type leaves the Droplet running without an expected disk and emits no diagnostic. Add dropletVolumesFromConfig that returns an explicit error for any non-string or empty entry, mirroring dropletSSHKeysFromConfig's error-message style. Tests cover both shapes: []any{123, "data"} (type mismatch) and []any{""} (empty string). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(volume): reject fractional size_gb instead of silent truncation (Copilot finding #2) intFromConfig truncates float64 silently — size_gb: 100.9 (e.g. via JSON or YAML float coercion) was creating a 100 GB volume with no diagnostic. Add intStrictFromConfig that mirrors the float64 fractional-rejection logic already in dropletSSHKeysFromConfig and use it for size_gb in Create, Update, and Diff. Whole-valued float64 (the structpb wire shape) still passes through. intFromConfig is unchanged for callers where rounding is acceptable (node_count, instance_count, probe seconds, etc.). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(droplet): accept []int and []int64 ssh_keys at top level (Copilot finding #3) dropletSSHKeysFromConfig accepted []any and []string at the top level but bailed on Go-native []int{101} or []int64{101} with "expected list, got []int" — even though numeric IDs are documented as supported and the per-element handler already covers int / int64 / float64. Add type-switches that mirror the per-element validation (non-positive rejection, error message style). Tests cover three shapes: []int{101,102}, []int64{555,666}, and the non-positive rejection path []int{0,101}. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(sizing): cover infra.volume in resolveSizing test cases (Copilot finding #4) infra.volume was added to doSizingMap as noopSizing but the manually- maintained TestResolveSizing cases table was not updated. Without the case, the registration is unverified and a future refactor that drops the entry would not fail this test. Add SizeS + SizeXL coverage so the "n/a" contract is locked in alongside every other noop-sized resource. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(volume): treat description/tags changes as ForceNew (Copilot finding #5) Diff previously tracked only size_gb / region / filesystem_type. Changes to description and tags were silently ignored because godo.StorageActions has no update endpoint for either, and the godo Volume API itself sets both at creation time only — so there is no in-place update path to make available. Resolution: surface drift as ForceNew (matching the way region / filesystem shrinks are handled) so the planner emits a planned replace rather than dropping the change. Document the constraint in the diff comment. Surface description and tags in volumeOutput so Diff has values to compare. Reuse outputsAsStringSlice + equalStringSet (firewall.go) so reordered tags do NOT trigger a needless replace. Tests: tags-add forces replace; reordered tags do NOT; description change forces replace. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(droplet): Diff flags vpc/tags/volumes/backups drift as ForceNew (Copilot finding #6) Extended droplet fields (user_data, vpc_uuid, ssh_keys, volumes, tags, enable_backups, monitoring, ipv6) were wired into Create only; Diff compared just "size", so changing any of these on an existing droplet produced no plan action and the drift was silently dropped. godo explicitly disallows Update on droplets (PUT only resizes), so Diff is the only place to flag these — every detected change must be ForceNew. Surface the comparable fields in dropletOutput (vpc_uuid, enable_backups derived from BackupIDs, tags, volumes) so Diff has values to compare; extend Diff with order-irrelevant set comparison for tags / volumes (reuses outputsAsStringSlice + equalStringSet from firewall.go). user_data, monitoring, ipv6, ssh_keys are NOT drift-checked: godo.Droplet does not surface them on Read, so comparing against current Outputs would produce a perpetually-dirty plan. Documented as a known limitation. Tests: vpc_uuid change forces replace; tags add forces replace; tags reorder does NOT force replace; enable_backups toggle forces replace. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(plugin): include 'Block Storage volumes' in Manifest.Description Manifest.Description must match plugin.json's description (test enforced in plugin_test.go). The merge from main pulled in a longer description that didn't yet mention Block Storage volumes; this aligns both. * fix(ci): wfctl-strict-contracts must copy plugin.contracts.json alongside plugin.json The Validate strict plugin contracts step (added in PR #41) copies only plugin.json to a tempfile when simulating GoReleaser's before-hook version rewrite. But wfctl plugin validate --strict-contracts looks for plugin.contracts.json CO-LOCATED with plugin.json (same directory). The rewrite-validation therefore always fails with "missing_module_contract_descriptor: module type \"iac.provider\" has no strict contract descriptor" — even when plugin.contracts.json IS valid in the repo root. This was masked on main because main has a single configurable contract file and any change rebuilds in-place; surfaced now by PR #55 which introduces a new resource type and exercises the second validation path. Fix: use a temp DIR and copy both plugin.json + plugin.contracts.json into it so the rewritten plugin.json's neighbour is intact. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(droplet): resolve VolumeIDs to names in Outputs (Copilot round-2 finding #1) dropletOutput previously stored godo's raw VolumeIDs (UUIDs) in Outputs["volumes"], while desired config carries volume *names*. After every successful Create/Read, the next Diff would compare e.g. ["vol-uuid-1"] against ["pg-data"] and force-replace the Droplet — catastrophic for stateful workloads (the PG data Droplet would be destroyed and recreated on every deploy, losing all game state). Resolve each VolumeID to its name via Storage.GetVolume in dropletOutput. Cache lookups within a single Read so duplicate VolumeIDs only hit the API once. On resolution failure (e.g. volume deleted out-of-band) fall back to the raw ID and record the unresolved IDs in Outputs["volumes_resolution"] so operators can debug. Update Diff comment to reflect that comparison is now name-vs-name and stable across plans. Tests: - Read with VolumeIDs returns names in Outputs - Diff with matching names returns no change (regression guard) - Diff with desired missing a name returns ForceNew - Resolution failure falls back to ID + records unresolved_ids - Duplicate IDs only call GetVolume once (cache hit) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(util): intStrictFromConfig surfaces type mismatches explicitly (Copilot round-2 finding #2) Previously `intStrictFromConfig` returned (defaultVal, false, nil) when the config value was an unsupported type (e.g. `size_gb: "100"` as a string). The caller then emitted "size_gb is required" — a misleading diagnostic that hid the real bug (operator quoted a number). Return present=true with an explicit "expected integer, got <type>" error so the caller surfaces the type problem directly. Tests: string-typed size_gb and bool-typed size_gb both reject with "expected integer, got <type>" before any CreateVolume call is made. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(droplet): tags clear (non-empty -> empty) surfaces as drift (Copilot round-2 finding #3) The Diff path skipped tag comparison when len(desired.tags)==0, so clearing tags on a Droplet was silently ignored. Operators sometimes need to strip tags to remove a Droplet from a tag-based firewall or backup schedule; "no diff" is dangerously wrong here. Switch the guard to "key present in desired" so empty desired vs non-empty current surfaces as ForceNew, while absent desired still skips (preserves backwards-compat for YAML that predates the field). Tests: clearing tags forces replace; absent tags key does NOT trigger drift; reorder still ignored. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(volume): filesystem_type transitions raw<->ext4 surface as drift (Copilot round-2 finding #4) The empty-side guard skipped raw(empty)->ext4 and ext4->raw transitions, so changing filesystem_type after Create produced no diff. DO Block Storage cannot reformat a volume in place, so any change must surface as ForceNew rather than being silently ignored. Switch to "key present in desired" guard so empty<->non-empty surfaces as drift. Absent desired key still skips (backwards-compat for YAML predating the field). Tests: raw->ext4 and ext4->raw both force replace; absent key does NOT trigger drift. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(volume): description add/clear from empty surfaces as drift (Copilot round-2 finding #5) The empty-side guard skipped both empty->non-empty (adding a description after Create) and non-empty->empty (clearing a description) transitions. DO Block Storage exposes no description-update endpoint, so any change must surface as ForceNew rather than being silently ignored. Switch to "key present in desired" guard. Absent key still skips (backwards-compat). Tests: add-from-empty and clear-to-empty both force replace; absent key does NOT trigger drift. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(volume): tags clear surfaces as drift (Copilot round-2 finding #6) Same pattern as droplet finding #3 and volume description finding #5. Clearing tags (non-empty current -> empty desired) was silently ignored because Diff skipped when len(desired.tags)==0. DO Block Storage has no tag-update endpoint, so any tag change must surface as ForceNew. Switch to "key present in desired" guard. Absent key still skips. Tests: clearing tags forces replace; absent tags key does NOT trigger drift (existing reorder/equal-set behavior preserved). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(droplet): vpc_uuid add-from-empty surfaces as drift (Copilot round-2 finding #7) The curVPC != "" guard skipped vpc_uuid drift when the current state had no vpc_uuid. Pre-release Droplet states (created before vpc_uuid became part of Outputs) would silently ignore an operator adding a vpc_uuid pin to YAML — exactly when explicit drift detection matters most. Switch to "key present in desired" guard. Empty current vs non-empty desired now triggers ForceNew. Absent desired key still skips (backwards-compat). Tests: vpc_uuid add-from-empty forces replace; absent key does NOT trigger drift. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * docs(droplet): explicit limitation block for user_data/ssh_keys/monitoring/ipv6 (Copilot round-2 finding #8) Round-1 documented the gap inline but the silent fallthrough at the end of Diff was easy to miss. Replace with an explicit block that: - Names each undetected field and the godo reason - States the operator workaround (`taint` / delete + re-apply) - References follow-up issue #56 for the resolution path - Updates the function-level Diff doc-comment to point at the inline rationale Issue: #56 "Droplet Diff misses user_data / ssh_keys / monitoring / ipv6 (godo Read limitation)" Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

PR #41 added a top-level `after:` hook to .goreleaser.yaml. goreleaser v2 has no top-level `after:` field — only `before:` exists. The release workflow for v0.9.0 (run 25274816649) failed instantly with `yaml: unmarshal errors: line 12: field after not found in type config.Project`. The previous `mv plugin.json.orig plugin.json` was theatre anyway: the GH Actions runner is ephemeral, so the in-place sed mutations to plugin.json are discarded when the runner shuts down. No need to restore plugin.json on the runner. Removed the matching `cp plugin.json plugin.json.orig` from the before block as well — it was preserving state nothing else reads. Validated with goreleaser/goreleaser:v2.15.4 check. Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…ession) (#59) wfctl bootstrap uses `provider: digitalocean` as the discriminator to identify which iac.provider module owns a given backend (e.g. `backend: spaces` for IaC state). v0.9.0's strict-contracts proto (PR #41) didn't declare `provider`, so configs carrying the canonical discriminator (BMW, core-dump, every existing iac.provider config) failed typed-config marshal with `protobuf error` because wfctl uses `protojson UnmarshalOptions{DiscardUnknown: false}` (workflow plugin/external/convert.go:65) which rejects unknown fields. The plugin itself doesn't use this field — manifest.iacProvider.name = digitalocean is the source of truth for provider identity — but the proto must accept it so strict-contracts unmarshal doesn't reject the config. Surfaced by core-dump's first deploy attempt against v0.9.0 (run 25277359916: `error: no iac.provider module found in config — add an iac.provider module to bootstrap remote state backends (backend="spaces")` after temporarily removing `provider:` to bypass the protobuf-error rejection). Bumps to v0.9.1. Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Initial plan

1d2ac9a

Copilot AI assigned Copilot and intel352 May 1, 2026

Copilot AI requested review from Copilot and removed request for Copilot May 1, 2026 06:12

intel352 marked this pull request as ready for review May 1, 2026 06:24

Copilot AI review requested due to automatic review settings May 1, 2026 06:24

Copilot AI reviewed May 1, 2026

View reviewed changes

Copilot started work on behalf of intel352 May 1, 2026 06:39 View session

Copilot AI linked an issue May 1, 2026 that may be closed by this pull request

Migrate workflow-plugin-digitalocean to strict gRPC proto contracts #40

Closed

Copilot AI and others added 2 commits May 1, 2026 06:49

chore: initial plan for strict gRPC proto contracts migration

7e3dbeb

Agent-Logs-Url: https://github.com/GoCodeAlone/workflow-plugin-digitalocean/sessions/7ff39de7-6d6d-4104-be5b-5685d34865c5 Co-authored-by: intel352 <77607+intel352@users.noreply.github.com>

Copilot AI requested review from Copilot and removed request for Copilot May 1, 2026 07:04

Copilot AI changed the title ~~[WIP] Migrate workflow-plugin-digitalocean to strict gRPC proto contracts~~ feat(contracts): migrate iac.provider to strict gRPC proto contracts May 1, 2026

Copilot AI requested a review from intel352 May 1, 2026 07:06

Copilot finished work on behalf of intel352 May 1, 2026 07:06

Copilot started work on behalf of intel352 May 1, 2026 08:24 View session

Copilot started work on behalf of intel352 May 1, 2026 12:46 View session

ci: add wfctl-strict-contracts validation job to ci.yml

5db3a02

Agent-Logs-Url: https://github.com/GoCodeAlone/workflow-plugin-digitalocean/sessions/f5a75b7b-047c-4e2c-954d-70bb922b6e07 Co-authored-by: intel352 <77607+intel352@users.noreply.github.com>

Copilot AI requested review from Copilot and removed request for Copilot May 1, 2026 12:51

github-advanced-security AI found potential problems May 1, 2026

View reviewed changes

Comment thread .github/workflows/ci.yml Fixed

ci: add explicit contents:read permissions to CI jobs

49d0f68

Agent-Logs-Url: https://github.com/GoCodeAlone/workflow-plugin-digitalocean/sessions/f5a75b7b-047c-4e2c-954d-70bb922b6e07 Co-authored-by: intel352 <77607+intel352@users.noreply.github.com>

Copilot AI requested review from Copilot and removed request for Copilot May 1, 2026 12:51

fix: tighten digitalocean strict review coverage

08d0d9e

intel352 requested a review from Copilot May 2, 2026 22:58

Copilot started reviewing on behalf of intel352 May 2, 2026 22:59 View session

Copilot AI reviewed May 2, 2026

View reviewed changes

Comment thread proto/generate.go Outdated

Comment thread internal/plugin_test.go Outdated

Comment thread .goreleaser.yaml Outdated

fix: address digitalocean package review

f97dec3

intel352 requested a review from Copilot May 2, 2026 23:27

Copilot started reviewing on behalf of intel352 May 2, 2026 23:27 View session

Copilot AI reviewed May 2, 2026

View reviewed changes

Comment thread plugin.json

Comment thread internal/plugin_test.go Outdated

Comment thread .goreleaser.yaml

Comment thread internal/plugin_test.go Outdated

Comment thread internal/plugin_test.go

Comment thread internal/plugin_test.go

fix: address digitalocean release archive review

93673af

intel352 requested a review from Copilot May 2, 2026 23:39

Copilot started reviewing on behalf of intel352 May 2, 2026 23:40 View session

Copilot AI reviewed May 2, 2026

View reviewed changes

Copilot started work on behalf of intel352 May 3, 2026 01:25 View session

Copilot AI and others added 2 commits May 3, 2026 01:33

fix: use portable sed -i.bak form in CI rewrite-simulation step

945ba40

Agent-Logs-Url: https://github.com/GoCodeAlone/workflow-plugin-digitalocean/sessions/55049097-392b-4939-aa13-b98c931fdb0f Co-authored-by: intel352 <77607+intel352@users.noreply.github.com>

Copilot AI requested review from Copilot and removed request for Copilot May 3, 2026 01:35

Copilot finished work on behalf of intel352 May 3, 2026 01:36

intel352 merged commit bbc8195 into main May 3, 2026
2 checks passed

intel352 mentioned this pull request May 3, 2026

fix(release): remove top-level after: hook (unblocks goreleaser v2) #57

Merged

2 tasks

intel352 mentioned this pull request May 3, 2026

fix(proto): accept provider field in IacProviderConfig (v0.9.0 regression) #59

Merged

4 tasks

		"version": "0.8.5",
		"description": "DigitalOcean IaC provider: App Platform, DOKS, databases, Redis cache, load balancers, VPC, firewall, DNS, Spaces, DOCR, certificates, Droplets, IAM (declared), and API gateway",

		if os.Getenv("WORKFLOW_PLUGIN_DIGITALOCEAN_DISABLE_LEGACY_MODULE") == "1" {
		return nil, fmt.Errorf("digitalocean plugin: legacy module path disabled")

		@@ -0,0 +1,3 @@
		package proto

		//go:generate sh -c "cd .. && protoc --go_out=. --go_opt=paths=source_relative proto/digitalocean.proto"

Conversation

Copilot AI commented May 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes

Validation

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

intel352 commented May 1, 2026

Uh oh!

intel352 commented May 1, 2026

Uh oh!

intel352 commented May 1, 2026

Uh oh!

Uh oh!

intel352 commented May 2, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

intel352 commented May 2, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

intel352 commented May 2, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

intel352 commented May 3, 2026

Uh oh!

Copilot AI commented May 3, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Copilot AI commented May 1, 2026 •

edited

Loading