apify-test-tools 0.10.2-beta.1 → 0.10.2-beta.2

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (70) hide show
  1. package/.gitattributes +2 -0
  2. package/.github/actions/checkout-restore-dependencies/action.yaml +13 -29
  3. package/.github/cli/apify-test-tools.mjs +362 -0
  4. package/.github/scripts/run-with-apify-tokens.mjs +2 -2
  5. package/.github/workflows/_check_code.yaml +25 -0
  6. package/.github/workflows/_move_major_tag.yaml +3 -81
  7. package/.github/workflows/_tests.yaml +34 -0
  8. package/.github/workflows/_update_release_metadata.yaml +0 -12
  9. package/.github/workflows/manual_move_major_tag.yaml +3 -3
  10. package/.github/workflows/manual_release_stable.yaml +3 -13
  11. package/.github/workflows/on_master.yaml +9 -5
  12. package/.github/workflows/on_pull_request.yaml +3 -48
  13. package/.github/workflows/public_platform-tests.yaml +1 -1
  14. package/.github/workflows/public_pr-build-test.yaml +2 -2
  15. package/.github/workflows/public_push-build-latest.yaml +2 -2
  16. package/.husky/pre-commit +5 -0
  17. package/.prettierignore +3 -0
  18. package/CHANGELOG.md +4 -0
  19. package/CONTRIBUTING.md +8 -18
  20. package/README.md +39 -44
  21. package/bin/build-from-local.ts +47 -13
  22. package/bin/build.ts +35 -19
  23. package/bin/main.ts +20 -5
  24. package/bin/slack.ts +28 -6
  25. package/bin/test-report.ts +6 -9
  26. package/bin/types.ts +9 -0
  27. package/dist/bin/build-from-local.d.ts +10 -2
  28. package/dist/bin/build-from-local.d.ts.map +1 -1
  29. package/dist/bin/build-from-local.js +29 -11
  30. package/dist/bin/build-from-local.js.map +1 -1
  31. package/dist/bin/build.d.ts +15 -5
  32. package/dist/bin/build.d.ts.map +1 -1
  33. package/dist/bin/build.js +24 -15
  34. package/dist/bin/build.js.map +1 -1
  35. package/dist/bin/main.js +18 -4
  36. package/dist/bin/main.js.map +1 -1
  37. package/dist/bin/slack.d.ts +8 -1
  38. package/dist/bin/slack.d.ts.map +1 -1
  39. package/dist/bin/slack.js +26 -5
  40. package/dist/bin/slack.js.map +1 -1
  41. package/dist/bin/test-report.d.ts.map +1 -1
  42. package/dist/bin/test-report.js +6 -8
  43. package/dist/bin/test-report.js.map +1 -1
  44. package/dist/bin/types.d.ts +8 -0
  45. package/dist/bin/types.d.ts.map +1 -1
  46. package/dist/test/bundle/cli-bundle.test.d.ts +2 -0
  47. package/dist/test/bundle/cli-bundle.test.d.ts.map +1 -0
  48. package/dist/test/bundle/cli-bundle.test.js +162 -0
  49. package/dist/test/bundle/cli-bundle.test.js.map +1 -0
  50. package/dist/test/e2e/cli-bundle.e2e.test.d.ts +2 -0
  51. package/dist/test/e2e/cli-bundle.e2e.test.d.ts.map +1 -0
  52. package/dist/test/e2e/cli-bundle.e2e.test.js +172 -0
  53. package/dist/test/e2e/cli-bundle.e2e.test.js.map +1 -0
  54. package/dist/test/unit/bin/build-from-local.test.js +53 -1
  55. package/dist/test/unit/bin/build-from-local.test.js.map +1 -1
  56. package/dist/test/unit/bin/build.test.js +113 -3
  57. package/dist/test/unit/bin/build.test.js.map +1 -1
  58. package/dist/test/unit/bin/slack.test.d.ts +2 -0
  59. package/dist/test/unit/bin/slack.test.d.ts.map +1 -0
  60. package/dist/test/unit/bin/slack.test.js +54 -0
  61. package/dist/test/unit/bin/slack.test.js.map +1 -0
  62. package/dist/tsconfig.tsbuildinfo +1 -1
  63. package/package.json +13 -3
  64. package/test/bundle/cli-bundle.test.ts +192 -0
  65. package/test/e2e/cli-bundle.e2e.test.ts +237 -0
  66. package/test/unit/bin/build-from-local.test.ts +55 -0
  67. package/test/unit/bin/build.test.ts +135 -3
  68. package/test/unit/bin/slack.test.ts +65 -0
  69. package/.github/scripts/check-package-version-bump.mjs +0 -74
  70. package/.github/workflows-package-version +0 -1
@@ -1,12 +1,12 @@
1
1
  #!/usr/bin/env node
2
2
  // Runs apify-test-tools and passes it only secrets needed for the Actors, no other secrets are leaked.
3
3
  //
4
- // node .github/scripts/run-with-apify-tokens.mjs npx apify-test-tools build --target-branch ...
4
+ // node .github/scripts/run-with-apify-tokens.mjs apify-test-tools build --target-branch ...
5
5
  //
6
6
  // The step passes the whole secrets map as ALL_SECRETS. The token names come from `tokenEnvVar` in
7
7
  // the repo's apify-test-tools.config.json, which is the same file apify-test-tools itself reads, so
8
8
  // the two can't drift. Everything else in the secrets map (npm, Slack, GitHub, anything else the
9
- // repo holds) is left out, and ALL_SECRETS itself is dropped, so the blob never reaches npx or
9
+ // repo holds) is left out, and ALL_SECRETS itself is dropped, so the blob never reaches the CLI or
10
10
  // anything under node_modules.
11
11
  //
12
12
  // The command runs without a shell, so branch names and other interpolated arguments are passed
@@ -92,3 +92,28 @@ jobs:
92
92
  run: npm ci
93
93
  - name: Build
94
94
  run: npm run build
95
+
96
+ # The composite action runs the CLI from .github/cli/apify-test-tools.mjs, which the pre-commit
97
+ # hook rebuilds. This catches commits that skipped the hook, so a stale bundle never ships.
98
+ cli_bundle:
99
+ name: CLI bundle
100
+ runs-on: ubuntu-latest
101
+ steps:
102
+ - name: Checkout repository
103
+ uses: actions/checkout@v6
104
+ - name: Use Node.js
105
+ uses: actions/setup-node@v6
106
+ with:
107
+ node-version: 24
108
+ cache: 'npm'
109
+ cache-dependency-path: 'package-lock.json'
110
+ - name: Install dependencies
111
+ run: npm ci
112
+ - name: Check the bundle is up to date
113
+ run: |
114
+ npm run bundle
115
+ if ! git diff --quiet -- .github/cli/; then
116
+ echo "::error::.github/cli/apify-test-tools.mjs is out of date. Run \`npm run bundle\` and commit the result."
117
+ exit 1
118
+ fi
119
+ node .github/cli/apify-test-tools.mjs --help > /dev/null
@@ -1,13 +1,8 @@
1
1
  name: Move major version tag
2
2
 
3
- # Consumer repos track a floating major tag (`...@v0`), not `master`, so merging to master does not
4
- # ship anything by itself. This workflow is what ships: it moves that tag forward, but only once the
5
- # package version the workflows pin in .github/workflows-package-version is actually on npm.
6
- #
7
- # That gate is the whole point. The workflows and the npm package release on their own schedules, so
8
- # the only combination that can break consumers is a workflow that calls a CLI feature which has not
9
- # been published yet. In that case the tag simply stays where it is and production keeps running the
10
- # previous workflows until someone cuts a release.
3
+ # Consumer repos track a floating major tag (`...@workflows-v0`), not `master`. This workflow is what
4
+ # ships: it moves that tag forward. The workflows and the CLI they call (bundled in .github/cli/)
5
+ # live in the same commit, so whatever the tag points at is consistent on its own.
11
6
 
12
7
  on:
13
8
  workflow_call:
@@ -16,13 +11,6 @@ on:
16
11
  description: Commit to move the tag to. Defaults to the commit that triggered the caller.
17
12
  type: string
18
13
  default: ''
19
- wait_for_publish:
20
- description: >-
21
- Set by a release that has just published the pinned version. npm accepts a publish
22
- before the version is readable, so the check waits instead of treating the gap as
23
- a held tag, and fails if the version never appears.
24
- type: boolean
25
- default: false
26
14
 
27
15
  permissions:
28
16
  contents: read
@@ -63,54 +51,7 @@ jobs:
63
51
  with:
64
52
  ref: ${{ inputs.ref || github.sha }}
65
53
 
66
- - name: Check the required package version is published
67
- id: check
68
- env:
69
- # npm accepts a publish and processes it asynchronously ("Your package is being
70
- # processed and may take a few minutes to become available"). Observed at ~19
71
- # minutes for a provenance-signed release, so the release path waits well past
72
- # npm's own estimate rather than shipping a tag consumers cannot install against.
73
- WAIT_SECONDS: ${{ inputs.wait_for_publish && 2400 || 0 }}
74
- run: |
75
- pinned_version=$(cat .github/workflows-package-version)
76
- echo "pinned_version=$pinned_version" >> "$GITHUB_OUTPUT"
77
-
78
- deadline=$(( $(date +%s) + WAIT_SECONDS ))
79
- while :; do
80
- # --prefer-online so a 404 cached from an earlier attempt cannot outlive the
81
- # publish it was polling for.
82
- if output=$(npm view "apify-test-tools@$pinned_version" version --prefer-online 2>&1); then
83
- echo "apify-test-tools@$pinned_version is on npm"
84
- echo "published=true" >> "$GITHUB_OUTPUT"
85
- exit 0
86
- fi
87
-
88
- # Anything that is not "this version does not exist" is a registry, network or
89
- # auth problem. Sitting in the retry loop would hide it behind a held tag.
90
- if ! grep -q 'code E404' <<< "$output"; then
91
- echo "::error::npm view failed for a reason other than the version being absent."
92
- echo "$output"
93
- exit 1
94
- fi
95
-
96
- if [ "$(date +%s)" -ge "$deadline" ]; then
97
- break
98
- fi
99
- echo "apify-test-tools@$pinned_version is not on npm yet, retrying in 30s"
100
- sleep 30
101
- done
102
-
103
- echo "published=false" >> "$GITHUB_OUTPUT"
104
-
105
- # A release waited for its own publish and never saw it. That is a broken release,
106
- # not a held tag, and it has to be red so it can be re-run once npm catches up.
107
- if [ "$WAIT_SECONDS" -gt 0 ]; then
108
- echo "::error::apify-test-tools@$pinned_version was published by this run but is still not readable after $((WAIT_SECONDS / 60))m. Re-run this job once it appears on npm."
109
- exit 1
110
- fi
111
-
112
54
  - name: Move tag
113
- if: steps.check.outputs.published == 'true'
114
55
  env:
115
56
  TARGET_SHA: ${{ inputs.ref || github.sha }}
116
57
  run: |
@@ -120,23 +61,4 @@ jobs:
120
61
  echo "### \`$MAJOR_TAG\` moved"
121
62
  echo
122
63
  echo "Now points at \`$TARGET_SHA\`."
123
- echo "Pins \`apify-test-tools@${{ steps.check.outputs.pinned_version }}\`, which is published."
124
- } >> "$GITHUB_STEP_SUMMARY"
125
-
126
- # Not a failure. Holding the tag is the designed outcome of merging a workflow change
127
- # that needs an unreleased package feature, and consumers are unaffected while it waits.
128
- - name: Report held tag
129
- if: steps.check.outputs.published != 'true'
130
- env:
131
- TARGET_SHA: ${{ inputs.ref || github.sha }}
132
- run: |
133
- {
134
- echo "### \`$MAJOR_TAG\` held"
135
- echo
136
- echo "\`$TARGET_SHA\` pins \`apify-test-tools@${{ steps.check.outputs.pinned_version }}\`, which is not on npm."
137
- echo "Consumers keep running the workflows \`$MAJOR_TAG\` already points at."
138
- echo
139
- echo "Publish that version with the **Stable release** workflow. The tag moves on the"
140
- echo "next master push, or immediately via the **Move major version tag** workflow."
141
64
  } >> "$GITHUB_STEP_SUMMARY"
142
- echo "::warning::$MAJOR_TAG not moved: apify-test-tools@${{ steps.check.outputs.pinned_version }} is not published yet."
@@ -3,6 +3,10 @@ name: Tests
3
3
  on:
4
4
  workflow_dispatch:
5
5
  workflow_call:
6
+ secrets:
7
+ # Token of an isolated sandbox account that can only reach the e2e test's two dummy Actors.
8
+ E2E_TEST_APIFY_TOKEN_DBYPODOIHUZABNQTR:
9
+ required: false
6
10
 
7
11
  permissions:
8
12
  contents: read
@@ -33,3 +37,33 @@ jobs:
33
37
 
34
38
  - name: Unit tests
35
39
  run: npm test
40
+
41
+ # Runs the shipped CLI bundle against real Actors (test/e2e). Node 24 only, so one push builds
42
+ # and runs the sandbox Actors once rather than once per Node version.
43
+ e2e_tests:
44
+ name: E2E tests
45
+ runs-on: ubuntu-latest
46
+ steps:
47
+ - name: Checkout repository
48
+ uses: actions/checkout@v6
49
+
50
+ - name: Use Node.js
51
+ uses: actions/setup-node@v6
52
+ with:
53
+ node-version: 24
54
+ cache: 'npm'
55
+ cache-dependency-path: 'package-lock.json'
56
+
57
+ - name: Install dependencies
58
+ run: npm ci
59
+
60
+ # The token is only on this step. Fork and Dependabot PRs don't get the secret, and the
61
+ # suite skips itself without it.
62
+ - name: E2E tests
63
+ env:
64
+ E2E_TEST_APIFY_TOKEN_DBYPODOIHUZABNQTR: ${{ secrets.E2E_TEST_APIFY_TOKEN_DBYPODOIHUZABNQTR }}
65
+ run: |
66
+ if [ -z "$E2E_TEST_APIFY_TOKEN_DBYPODOIHUZABNQTR" ]; then
67
+ echo "::notice::E2E_TEST_APIFY_TOKEN_DBYPODOIHUZABNQTR is not available to this run, the e2e tests are skipped."
68
+ fi
69
+ npm run test:e2e
@@ -77,18 +77,6 @@ jobs:
77
77
  write-mode: overwrite
78
78
  contents: ${{ needs.release_metadata.outputs.changelog }}
79
79
 
80
- # The reusable workflows install this exact version, so the commit that releases a
81
- # version is the commit that pins it. That is what lets a frozen major tag keep
82
- # installing the library it was tested against instead of drifting to whatever is
83
- # newest.
84
- #
85
- # Skipped for prereleases: master pushes release a `-beta` that consumer repos must
86
- # never install. Their pin stays on the last stable, and the beta stays reachable
87
- # through the lockfile path in the setup action, which is how branch testing works.
88
- - name: Update the version the workflows pin
89
- if: inputs.release_type != 'prerelease'
90
- run: echo "${{ needs.release_metadata.outputs.version_number }}" > .github/workflows-package-version
91
-
92
80
  - name: Commit changes
93
81
  id: commit
94
82
  uses: apify/actions/signed-commit@v1.0.0
@@ -1,8 +1,8 @@
1
1
  name: Move major version tag
2
2
 
3
- # Escape hatch for the two cases the automatic move does not cover: shipping a workflow change right
4
- # after the release that unblocked it, without waiting for the next master push, and rolling the tag
5
- # back to an earlier commit when a shipped workflow turns out to be broken.
3
+ # Escape hatch for what the automatic move on master pushes does not cover: rolling the tag back to
4
+ # an earlier commit when a shipped workflow or CLI turns out to be broken, or moving it after a push
5
+ # whose checks failed for an unrelated reason.
6
6
 
7
7
  on:
8
8
  workflow_dispatch:
@@ -36,6 +36,9 @@ jobs:
36
36
  tests:
37
37
  name: Tests
38
38
  uses: ./.github/workflows/_tests.yaml
39
+ # Only the sandbox token, not `secrets: inherit`: the test job runs test code.
40
+ secrets:
41
+ E2E_TEST_APIFY_TOKEN_DBYPODOIHUZABNQTR: ${{ secrets.E2E_TEST_APIFY_TOKEN_DBYPODOIHUZABNQTR }}
39
42
 
40
43
  release_metadata:
41
44
  name: Update release metadata
@@ -79,16 +82,3 @@ jobs:
79
82
  "ref": "${{ needs.release_metadata.outputs.changelog_commitish }}",
80
83
  "tag": "latest"
81
84
  }
82
-
83
- # This release is what unblocks a major tag being held for an unpublished package version, so
84
- # try the move again now rather than making someone wait for the next master push. A no-op when
85
- # the tag was not being held.
86
- move_major_tag:
87
- name: Move major version tag
88
- needs: [release_metadata, npm_publish]
89
- permissions:
90
- contents: write
91
- uses: ./.github/workflows/_move_major_tag.yaml
92
- with:
93
- ref: ${{ needs.release_metadata.outputs.changelog_commitish }}
94
- wait_for_publish: true
@@ -24,14 +24,18 @@ jobs:
24
24
  if: "!startsWith(github.event.head_commit.message, 'docs')"
25
25
  name: Tests
26
26
  uses: ./.github/workflows/_tests.yaml
27
+ # Only the sandbox token, not `secrets: inherit`: the test job runs test code.
28
+ secrets:
29
+ E2E_TEST_APIFY_TOKEN_DBYPODOIHUZABNQTR: ${{ secrets.E2E_TEST_APIFY_TOKEN_DBYPODOIHUZABNQTR }}
27
30
 
28
- # Ships any workflow change in this push to consumer repos. Deliberately independent of the npm
29
- # release below: a workflow-only change needs no release, and a package-only release leaves the
30
- # workflows byte-identical. `code_checks` runs actionlint, which is what validates the workflows;
31
- # `tests` covers the npm package and is skipped for docs commits, so it is not a dependency here.
31
+ # Ships this push to consumer repos: the workflows and the CLI bundled in .github/cli/ go live
32
+ # together. Independent of the npm release below, which only matters to people importing the lib.
33
+ # `code_checks` runs actionlint and checks the bundle is up to date. `tests` cover the CLI but are
34
+ # skipped for docs commits, so a skipped run does not hold the tag; a failed one does.
32
35
  move_major_tag:
33
36
  name: Move major version tag
34
- needs: [code_checks]
37
+ needs: [code_checks, tests]
38
+ if: ${{ !failure() && !cancelled() }}
35
39
  permissions:
36
40
  contents: write
37
41
  uses: ./.github/workflows/_move_major_tag.yaml
@@ -21,51 +21,6 @@ jobs:
21
21
  tests:
22
22
  name: Tests
23
23
  uses: ./.github/workflows/_tests.yaml
24
-
25
- # Catches the one combination that breaks consumers: a workflow starting to use a CLI feature
26
- # from the same PR. Merging ships the workflow immediately, so the pinned version has to move in
27
- # the same change or the workflow goes live calling something that is not published.
28
- package_version_bump_needed:
29
- name: Package version bump needed
30
- runs-on: ubuntu-latest
31
- steps:
32
- - name: Checkout repository
33
- uses: actions/checkout@v6
34
- with:
35
- fetch-depth: 0
36
- ref: ${{ github.event.pull_request.head.sha }}
37
-
38
- - name: Check whether this PR needs a version bump
39
- env:
40
- BASE_REF: ${{ github.base_ref }}
41
- PR_LABELS: ${{ join(github.event.pull_request.labels.*.name, ',') }}
42
- run: |
43
- git fetch --no-tags origin "$BASE_REF"
44
- git diff --name-only FETCH_HEAD...HEAD | node .github/scripts/check-package-version-bump.mjs
45
-
46
- # Reports whether merging this PR will ship the workflows or hold the major tag until a release.
47
- # Never fails: pinning a version ahead of the release that publishes it is a supported way to
48
- # land the package change and the workflow change that needs it in one PR.
49
- workflows_package_version:
50
- name: Workflows package version
51
- runs-on: ubuntu-latest
52
- steps:
53
- - name: Checkout repository
54
- uses: actions/checkout@v6
55
-
56
- - name: Check the pinned package version is published
57
- run: |
58
- pinned_version=$(cat .github/workflows-package-version)
59
- if npm view "apify-test-tools@$pinned_version" version >/dev/null 2>&1; then
60
- echo "Workflows pin apify-test-tools@$pinned_version, which is published."
61
- echo "The major tag moves on merge and these workflows go live." >> "$GITHUB_STEP_SUMMARY"
62
- else
63
- {
64
- echo "### Major tag will be held on merge"
65
- echo
66
- echo "These workflows pin \`apify-test-tools@$pinned_version\`, which is not on npm yet."
67
- echo "Merging is fine: consumers keep running the current workflows until you cut a"
68
- echo "stable release, and the tag moves once that version is published."
69
- } >> "$GITHUB_STEP_SUMMARY"
70
- echo "::notice::Merging will not ship these workflows until apify-test-tools $pinned_version is released."
71
- fi
24
+ # Only the sandbox token, not `secrets: inherit`: the test job runs test code.
25
+ secrets:
26
+ E2E_TEST_APIFY_TOKEN_DBYPODOIHUZABNQTR: ${{ secrets.E2E_TEST_APIFY_TOKEN_DBYPODOIHUZABNQTR }}
@@ -47,7 +47,7 @@ jobs:
47
47
  --fileParallelism=true \
48
48
  --maxWorkers 100
49
49
  return_code=$?
50
- npx apify-test-tools report-tests \
50
+ apify-test-tools report-tests \
51
51
  --report-file ./test-output.json \
52
52
  --workflow-name "${{ github.workflow }}" \
53
53
  --job-url "${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}" \
@@ -65,7 +65,7 @@ jobs:
65
65
  )
66
66
  [ -n "$base_commit" ] && args+=(--base-commit "$base_commit")
67
67
  actor_builds=$(node "$SCRIPTS_PATH/run-with-apify-tokens.mjs" \
68
- npx apify-test-tools build "${args[@]}")
68
+ apify-test-tools build "${args[@]}")
69
69
  echo "actor_builds=$actor_builds" | tee -a "$GITHUB_OUTPUT"
70
70
 
71
71
  # Runs the repo's test code and its whole import graph, so it gets the tester token only.
@@ -88,7 +88,7 @@ jobs:
88
88
  --target-branch "origin/$BASE_REF"
89
89
  )
90
90
  [ -n "$base_commit" ] && args+=(--base-commit "$base_commit")
91
- npx apify-test-tools get-latest-commit "${args[@]}" > ./base_commit.txt
91
+ apify-test-tools get-latest-commit "${args[@]}" > ./base_commit.txt
92
92
  echo "New last validated (base) commit: $(cat base_commit.txt)"
93
93
 
94
94
  # Pure static checks against the repo. Needs no Apify or Slack credentials, only the npm
@@ -36,11 +36,11 @@ jobs:
36
36
  SLACK_TOKEN_RELEASES_BOT: ${{ secrets.SLACK_TOKEN_RELEASES_BOT }}
37
37
  run: |
38
38
  node "${{ steps.setup.outputs.scripts-path }}/run-with-apify-tokens.mjs" \
39
- npx apify-test-tools release --base-commit ${{ github.event.before }} --release-slack-channel "#delivery-public-actors" --report-slack-channel "${{ inputs.report-slack-channel || format('#notif-{0}', github.event.repository.name) }}" --use-docker-cache
39
+ apify-test-tools release --base-commit ${{ github.event.before }} --release-slack-channel "#delivery-public-actors" --report-slack-channel "${{ inputs.report-slack-channel || format('#notif-{0}', github.event.repository.name) }}" --use-docker-cache
40
40
 
41
41
  - name: Delete old builds
42
42
  env:
43
43
  ALL_SECRETS: ${{ toJSON(secrets) }}
44
44
  run: |
45
45
  node "${{ steps.setup.outputs.scripts-path }}/run-with-apify-tokens.mjs" \
46
- npx apify-test-tools delete-old-builds
46
+ apify-test-tools delete-old-builds
package/.husky/pre-commit CHANGED
@@ -1,6 +1,11 @@
1
1
  echo "pre-commit: typechecking, linting, testing, checking code formatting, checking for unused exports & validating schemas"
2
2
 
3
3
 
4
+ # The workflows run the CLI from this bundle, so it ships with every commit that changes bin/ or its
5
+ # dependencies. Rebuilding takes well under a second.
6
+ npm run -s bundle
7
+ git add .github/cli/apify-test-tools.mjs
8
+
4
9
  npm test
5
10
  npm run format:check
6
11
  # We typecheck & lint last when we know our code works
package/.prettierignore CHANGED
@@ -8,3 +8,6 @@ CHANGELOG.md
8
8
 
9
9
  # Fixtures must stay byte-exact — several are deliberately malformed or empty
10
10
  test/fixtures
11
+
12
+ # Generated by `npm run bundle`
13
+ .github/cli/
package/CHANGELOG.md CHANGED
@@ -5,6 +5,10 @@ All notable changes to this project will be documented in this file.
5
5
  <!-- git-cliff-unreleased-start -->
6
6
  ## 0.10.2 - **not yet released**
7
7
 
8
+ ### 🚀 Features
9
+
10
+ - **ci:** Ship the workflows&#x27; CLI as a bundle in the action instead of installing it ([#200](https://github.com/apify/apify-test-tools/pull/200)) ([a0ef975](https://github.com/apify/apify-test-tools/commit/a0ef9756123a30a8210590b080a2a289c6a9c17c)) by [@metalwarrior665](https://github.com/metalwarrior665)
11
+
8
12
  ### 🐛 Bug Fixes
9
13
 
10
14
  - **toBeObject:** Reject null ([#191](https://github.com/apify/apify-test-tools/pull/191)) ([6bdae49](https://github.com/apify/apify-test-tools/commit/6bdae4905756e0ae8dd6146942e6fc04dbead48e)) by [@JuanGalilea](https://github.com/JuanGalilea), closes [#88](https://github.com/apify/apify-test-tools/issues/88)
package/CONTRIBUTING.md CHANGED
@@ -89,24 +89,14 @@ Consumers pin `@workflows-v0`, not `@master`. See
89
89
  [Versioning and releases](./README.md#versioning-and-releases) in the README for how the tag and the
90
90
  npm release relate — the short version:
91
91
 
92
- - changing only a workflow needs no npm release
93
- - changing only the package needs no workflow change
94
- - a workflow that calls a **new** CLI feature must set `.github/workflows-package-version` to the
95
- version that will contain it, in the same PR. The `workflows-v0` tag is then held until that version is on
96
- npm, so merging can't ship a workflow that calls a CLI that doesn't exist yet.
97
-
98
- `.github/workflows-package-version` holds one exact version, and the stable release writes it. The
99
- workflows install exactly it, which is what makes a frozen major tag stay frozen — it keeps the
100
- library it was tested with instead of following `latest` forever.
101
-
102
- The `Package version bump needed` check enforces the point above: a PR touching both a `public_`
103
- workflow (or the composite action) and `bin/`, `lib/` or `index.ts` has to move the pin. When the
104
- two changes are unrelated and the workflow doesn't need the new code, label the PR
105
- `no-version-bump-needed`.
106
-
107
- It only looks at a single PR, so it won't catch a workflow that starts using a CLI feature merged in
108
- an earlier, still-unreleased PR. That needs someone to land a CLI change and sit on it unreleased;
109
- catching it would mean flagging every workflow edit made while any package change is unreleased.
92
+ - the workflows and the CLI ship together on every master push, through the tag; no npm release needed
93
+ - the lib ships through a stable npm release
94
+ - the CLI runs from `.github/cli/apify-test-tools.mjs`, a bundle the pre-commit hook rebuilds with
95
+ `npm run bundle`. If you commit with `--no-verify`, run it yourself: the `CLI bundle` check fails
96
+ on a stale bundle, and a stale bundle would hold the tag.
97
+ - `npm test` covers the bundle without network (`test/bundle/`). `npm run test:e2e` runs it against two
98
+ dummy Actors in an isolated sandbox account (`test/e2e/`) and skips itself unless
99
+ `E2E_TEST_APIFY_TOKEN_DBYPODOIHUZABNQTR` is set. CI passes that secret only to the `E2E tests` job.
110
100
 
111
101
  `npm run lint` and `actionlint` (via the `Code checks` workflow) both gate master, so run them before
112
102
  pushing workflow changes.
package/README.md CHANGED
@@ -308,48 +308,35 @@ Slack credentials in scope, so a postinstall script in the dependency tree canno
308
308
 
309
309
  ## Versioning and releases
310
310
 
311
- The workflows and the npm package live in one repo but ship on their own schedules. Two pointers
312
- decide what a consumer repo actually runs:
313
-
314
- | Pointer | What it selects | Moves when |
315
- | ----------------------------------- | ---------------------------------------------- | -------------------------------------------------------- |
316
- | the `@workflows-v0` tag in `uses:` | which workflows run | a master push, once the version below is published |
317
- | `.github/workflows-package-version` | which `apify-test-tools` the workflows install | a stable release writes it; you may set it ahead of time |
318
-
319
- The setup action installs that **exact** version — not a range. A tag is therefore a complete
320
- statement: these workflows _and_ this library. The stable release writes the file into the same
321
- commit that bumps `package.json` and `CHANGELOG.md`, so a released commit always names the version
322
- it published, and rollout latency is unchanged: the release publishes and moves the tag in one run.
323
-
324
- Prereleases are excluded. Master pushes publish a `-beta` that consumer repos must never install, so
325
- the pin only moves on a stable release; betas stay reachable through the lockfile path in the setup
326
- action, which is how branch testing works.
327
-
328
- Nothing is coupled that doesn't need to be:
329
-
330
- - **Workflow-only change** — merge it. `workflows-v0` moves, it goes live, no release needed.
331
- - **Package-only change** — merge it, then cut a release when you want it out. The workflows are
332
- unchanged, so consumers see nothing until the release lands.
333
- - **A workflow that calls a new CLI feature** — the one case that can break consumers, and the only
334
- one with any ceremony. Put the package change, the workflow change, and the new pin in one PR. On
335
- merge the tag is **held**: the CI job reports that the pinned version isn't on npm and leaves `workflows-v0`
336
- where it is, so consumers keep running the previous workflows. Cut a stable release, and the tag
337
- moves on its own. Run **Move major version tag** if you don't want to wait for the next master
338
- push.
311
+ The workflows and the npm package live in one repo but ship on their own schedules:
312
+
313
+ | What | Consumers get it through | Ships when |
314
+ | -------------------------------------------- | ------------------------------------------ | ----------------------- |
315
+ | the workflows and the CLI they call | the `@workflows-v0` tag in `uses:` | every master push |
316
+ | the lib (`testActor`, `describe`, matchers…) | `apify-test-tools` in their `package.json` | a stable release on npm |
317
+
318
+ The CLI is bundled with its dependencies into `.github/cli/apify-test-tools.mjs` (`npm run bundle`,
319
+ run by the pre-commit hook and checked in CI), and the setup action runs it from there. Nothing is
320
+ installed for it, so a tag is a complete statement: these workflows and this CLI, with the exact
321
+ dependency versions from `package-lock.json` at that commit. The consumer's own `node_modules` is
322
+ left alone, so their tests keep the lib version they locked.
323
+
324
+ - **Workflow or CLI change** — merge it. `workflows-v0` moves once checks and tests pass, and it goes
325
+ live. No release needed, even when a workflow starts using a new CLI feature: both are in the
326
+ same commit.
327
+ - **Lib change** — merge it, then cut a release when you want it out. Consumers pick it up when
328
+ they update `apify-test-tools`.
329
+
330
+ The CLI and the lib in one job can be different versions now, so the formats between them are a
331
+ contract: the `ACTOR_BUILDS` JSON that `build` writes and the lib reads, and the vitest JSON report
332
+ `report-tests` reads. Add fields freely; renaming or removing one needs a new major tag.
339
333
 
340
334
  `workflows-v0` moving on every master push means `@workflows-v0` is as live as `@master` was —
341
- there's no staging step, just a gate on the package version. What the tag buys you is a
342
- `workflows-v1` for breaking workflow changes, so repos migrate one at a time instead of all at once,
343
- and a way to roll back by pointing the tag at an earlier commit. Bump `MAJOR_TAG` in
335
+ there's no staging step. What the tag buys you is a `workflows-v1` for breaking workflow changes,
336
+ so repos migrate one at a time instead of all at once, and a way to roll back by pointing the tag at
337
+ an earlier commit (**Move major version tag** workflow). Bump `MAJOR_TAG` in
344
338
  `.github/workflows/_move_major_tag.yaml` to cut the next major; the old tag then freezes where it is
345
- and keeps working.
346
-
347
- Freezing is real, which is the whole reason the version is pinned rather than a floor. A frozen `workflows-v0`
348
- points at a commit whose pinned version never changes, so it keeps installing the library it was
349
- tested against no matter how far the package moves on. Had the workflows installed `>=<floor>`, a
350
- frozen `workflows-v0` would still resolve to whatever is newest — and since the release that enables
351
- `workflows-v1` is usually the same release that breaks `workflows-v0`, the migration window would
352
- have been zero.
339
+ and keeps working, CLI included, since the bundle is part of the frozen commit.
353
340
 
354
341
  The tag tracks the **workflows'** contract, not the npm package version. They move independently on
355
342
  purpose, so `@workflows-v0` is expected to stay `@workflows-v0` after the package reaches 1.0 —
@@ -360,9 +347,8 @@ and those refs disagree, which is the mistake that would otherwise ship silently
360
347
  ### Testing workflow changes
361
348
 
362
349
  - Point [testing-repo-for-github-actions](https://github.com/apify-store/testing-repo-for-github-actions)
363
- at your branch (`uses: ...@your-branch`). It has real attached Actors and tests. Because the
364
- package lives here too, a master push publishes a `beta`, and the lockfile-beta path in the setup
365
- action installs that exact version — so one branch tests both halves of a change together.
350
+ at your branch (`uses: ...@your-branch`). It has real attached Actors and tests. The branch carries
351
+ its own CLI bundle, so one branch tests the workflows and the CLI together.
366
352
  - To change the composite action itself, repoint the `uses:` refs inside the reusable workflows at
367
353
  your branch as well, and change them back before merging.
368
354
  - Make sure the shell code actually works on your laptop first.
@@ -580,6 +566,15 @@ describe('standby tests', () => {
580
566
 
581
567
  The package includes a CLI binary used by CI workflows to build Actors, detect changes, and report test results. You can also run it locally.
582
568
 
569
+ ### Dry runs (`--dry-run`)
570
+
571
+ Every command that changes something (`build`, `build-from-local`, `release`, `delete-old-builds`, `report-tests`) takes `--dry-run`. A dry run:
572
+
573
+ - **Changes nothing.** No versions or builds are created, nothing is uploaded or deleted, and no Slack message is posted.
574
+ - **Does every read and check the real run does**, so it fails wherever the real run would: change detection, the Apify token and Actor lookup, the [repository check](#repository-check), collecting and zipping local sources, and the Slack token (verified with Slack's read-only `auth.test`).
575
+ - **Needs the same tokens as the real run.** Read-only tokens are enough, since nothing is written. A dry run can't tell whether a token is also allowed to write, because the Apify API has no dry-run mode of its own.
576
+ - **Prints a plan, not results.** Logs start with `[DRY RUN] Would …`. `build` and `build-from-local` write `{ "dryRun": true, "builds": [...] }` to stdout instead of the build array, so the plan can't be passed to tests as `ACTOR_BUILDS` by mistake.
577
+
583
578
  ### Running locally
584
579
 
585
580
  Running the testing library locally is useful when you only want to update the testing code in /test because you can iterate on it without pushing new code to the remote.
@@ -610,7 +605,7 @@ APIFY_TOKEN_JOHN_DOE=<token> \
610
605
  --dry-run
611
606
  ```
612
607
 
613
- Remove `--dry-run` to actually trigger builds and update the branch names/ The command outputs a JSON array of build objects to stdout:
608
+ With `--dry-run`, it prints the plan (the version and Git URL each Actor would be built from). Remove `--dry-run` to actually trigger builds and update the branch names. The command outputs a JSON array of build objects to stdout:
614
609
 
615
610
  ```json
616
611
  [{ "buildId": "...", "actorRawId": "...", "buildNumber": "...", "actorFullName": "john.doe/my-actor" }]
@@ -625,7 +620,7 @@ APIFY_TOKEN_JOHN_DOE=<token> \
625
620
  npx apify-test-tools build-from-local --actors john.doe/my-actor
626
621
  ```
627
622
 
628
- Pass a hardcoded actor name via `--actors` to build only that Actor (comma-separate multiple names). Omit `--actors` to build all Actors in the repo, or add `--dry-run` to preview without building. It outputs the same JSON build array as `build`, so you run tests against it the same way as in step 5 below:
623
+ Pass a hardcoded actor name via `--actors` to build only that Actor (comma-separate multiple names). Omit `--actors` to build all Actors in the repo, or add `--dry-run` to check the token, the Actor and the zipped sources without uploading or building. It outputs the same JSON build array as `build`, so you run tests against it the same way as in step 5 below:
629
624
 
630
625
  ```bash
631
626
  # Build from local source and capture output