Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
25 commits
Select commit Hold shift + click to select a range
72e0fb1
✨ hub: point owners to their spoke dashboard and the row ⋯ menu (#16589)
clubanderson Nov 4, 2026
8b0db18
🐛 fix(dashboard): a part-filled CI ring is a running CI chevron (#16632)
clubanderson Nov 4, 2026
73ccd4b
🐛 Activity Monitor: re-review waits on the hive reviewer; CI from a f…
clubanderson Nov 4, 2026
09c6d46
fix(dashboard): place the base-branch pill beside the repo chip (#16621)
clubanderson Nov 4, 2026
f20a228
🐛 ACMM: recognize install and boot E2E workflows (#16640)
clubanderson Nov 4, 2026
da691ee
🐛 Activity Monitor: say 'hive bot lacks permission', not 'unauthorize…
clubanderson Nov 4, 2026
655fa3f
fix(automerge): never supersede a hive-authored PR for a missing sign…
clubanderson Nov 4, 2026
bc23607
🐛 PR in progress chevron defers to running CI/Review and never speaks…
clubanderson Nov 4, 2026
01c5727
🐛 fix: require sign-off for workflow review questions (#16646)
Karibhou Nov 4, 2026
df66e91
fix: make merged ledger writes race-safe (#16633)
hivecommons-hive[bot] Nov 4, 2026
2a16238
fix(dashboard): accept triage advances the issue to 010 actionable (#…
clubanderson Nov 4, 2026
5c13b4f
🐛 fix(hub): back off all commit-message fetches on a throttled 403/42…
hivecommons-hive[bot] Nov 4, 2026
718eb7e
🌱 Activity Monitor: one chevron word, Queued for merge, for every PR …
clubanderson Nov 4, 2026
6366036
🐛 fix(dashboard): failed SSO handoff shows the sign-in page instead o…
hivecommons-hive[bot] Nov 4, 2026
ab12e49
🐛 fix: classify sentinel pacing skips (#16645)
Karibhou Nov 4, 2026
8ff1652
🐛 fix(dashboard): Review and CI chevrons state what the head waits on…
clubanderson Nov 4, 2026
9408464
🌱 chore(dashboard): share one relative-age formatter in tracker-row-s…
hivecommons-hive[bot] Nov 4, 2026
0a7212e
🐛 Activity Monitor: celebrate externally merged PRs with bee + toast …
clubanderson Nov 4, 2026
8b824c4
🐛 Stage tracker: say CI running (not activity unknown) when checks ar…
clubanderson Nov 4, 2026
859b7df
📖 docs(api-reference): list check-count and perspective fields on act…
hivecommons-hive[bot] Nov 4, 2026
04e5ca0
📖 docs: align StageTrackerActivity CI provenance comment (#16674)
thatguywob Nov 5, 2026
f358b73
🌱 test(crux): pin that a bad OTEL config disables telemetry, not the …
hivecommons-hive[bot] Nov 5, 2026
dcc92db
🐛 fix(ci): retry transient Cline dependency installs (#16667)
mendezr Nov 5, 2026
8565377
🐛 review: superseding PR with unanimous sign-off requires_human parks…
clubanderson Nov 5, 2026
fccafaa
Merge remote-tracking branch 'origin/v5' into topup/v6-from-v5-202611…
clubanderson Nov 5, 2026
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions changelog.d/added-hub-spoke-pointer.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Hub My Hives: a dismissible "Your hive <name> is ready → Open your hive ↗" callout for owners explains that the hub only lists hives (the Guide, settings and GitHub App install live inside the hive), names any attention item that must be fixed there, and points at the row ⋯ menu; the hive-name link gains an ↗ glyph and tooltip and the ⋮ button a "Manage: permissions, pause, funding, delete…" hint. Dismissal is per user and hive in localStorage and re-arms when a new attention item appears.
1 change: 1 addition & 0 deletions changelog.d/changed-16668-shared-age-word.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- The stage tracker's row tooltip and current-stage line now share one relative-age formatter (#16668)
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Activity Monitor rows now draw a PR's base-branch pill (`v5`, `main`) right after the repo chip at the head of the row, before the issue number, instead of in the right-hand modifier cluster; the pill names its PR on hover ("PR #N targets v5") and rows with no linked PR show none. ([#16620](https://github.com/hivecommons/hive/issues/16620))
1 change: 1 addition & 0 deletions changelog.d/changed-hive-permission-pill-wording.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Activity Monitor: the muted pill a PR wears while the hive's GitHub App is missing a permission now reads "hive bot lacks permission" instead of "unauthorized", and its tooltip, the row hover line, the Needs-attention chip title, the needs-human tip, the human-merge gate word and the decision popover say the same — nobody did anything illegitimate, the App installation lacks one permission (hivecommons/hive#16639). The `hive:unauthorized` label, comment marker and API fields are unchanged.
1 change: 1 addition & 0 deletions changelog.d/changed-queued-for-merge-wording.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Activity Monitor: every PR whose only remaining step is the merge itself now reads one chevron word, **Queued for merge**, with a short qualifier after the dot saying why it has not merged yet — *· next pass in 42s*, *· checking now*, *· soak 12m left*, *· 2nd of 3* (merge train), *· next pass* (Approve / Let hive decide), *· after CI* (Approve and let hive merge while checks run, or a merge call refused for unreported checks), *· needs rebase* (head moved under the merge call), *· branch protection pending*. Three adjacent approved rows on hive's own tracker read *Ready for auto merge …*, *Approved · hive will…* and *Approved · merge queu…*; the tooltip keeps the full explanation. *Soaking*, *Merging*, *Rebasing*, *Needs human decision* and a *Merge failed* a person must fix stay their own words (hivecommons/hive#16653).
1 change: 1 addition & 0 deletions changelog.d/fixed-16629.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Merged ledger writes no longer fail with "rename ... no such file or directory" when two writers race on the same temp file (#16629).
1 change: 1 addition & 0 deletions changelog.d/fixed-16649.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Hub SHA poller backs off all commit-message fetches on a 403/429 and warns once per throttle window instead of once per SHA, so a rate-limited token no longer trips the `sha_poll_commit_message_unavailable` fleet report ([#16649](https://github.com/hivecommons/hive/issues/16649)).
1 change: 1 addition & 0 deletions changelog.d/fixed-16652-sso-failure-shows-login.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Opening a hosted hive from the hub when the hive cannot verify the hub's single sign-on now shows the "Sign-in didn't complete" page with a GitHub sign-in button, instead of being replaced by the gateway's "Hive is restarting" outage page ([#16652](https://github.com/hivecommons/hive/issues/16652)).
1 change: 1 addition & 0 deletions changelog.d/fixed-16670-api-reference-activity-fields.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- API reference now lists the check-count and perspective fields of stage-tracker activity observations ([#16670](https://github.com/hivecommons/hive/issues/16670)).
1 change: 1 addition & 0 deletions changelog.d/fixed-16674-stage-tracker-ci-comment.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Clarify the `StageTrackerActivity` source documentation: CI activity uses the resolved required checks when branch protection is known, otherwise all check runs observed on the head. This corrects the description without changing runtime behavior ([#16674](https://github.com/hivecommons/hive/pull/16674)).
1 change: 1 addition & 0 deletions changelog.d/fixed-accept-triage-advances-stage.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Activity Monitor "Accept triage" now moves the issue to 010 Actionable in the same request: the handler writes `stage/010-actionable` alongside `triage/accepted` when the issue has no stage label or wears `stage/015-triage-needed` (never regressing a later stage, never on a held issue), and the tracker no longer reads an issue carrying `triage/accepted` with no stage label as 000-new while waiting for the engine's next pass.
1 change: 1 addition & 0 deletions changelog.d/fixed-acmm-install-e2e.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Recognize install, boot, and explicitly named E2E GitHub Actions workflows in the ACMM E2E prerequisite, retaining web/configuration and directory alternatives without treating unrelated CI workflows as E2E coverage.
1 change: 1 addition & 0 deletions changelog.d/fixed-ci-activity-unknown-caption.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Activity Monitor stage tracker: a PR head whose checks are still reporting now reads `CI running · 4 checks pending` (chevron `CI 7/11 · 4 running`, or the last one or two unfinished names) instead of `CI activity unknown · last observed missing` / `CI unknown · no check results observed` / `no verdict yet`; a required check with no run while others still execute is one sentence in the tooltip rather than the "missing" verdict; the "no check results observed" wording is reserved for a head with zero check runs; and the chevron tooltip states each fact once (count word, detail sentence, head, source, observation time) instead of repeating the unreported check four times (#16630). Check names embedded in captions and chevron text are wrapped in single quotes.
1 change: 1 addition & 0 deletions changelog.d/fixed-ci-partial-ring-running.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Activity Monitor: a CI ring that is not yet full now renders the CI chevron active ("CI 13/19 · running") beside a running Review instead of a silent part-green glyph; check-run motion on the head leads red, unreported or ambiguous neighbours, and icon-only green is reserved for a complete ring (hivecommons/hive#16631).
1 change: 1 addition & 0 deletions changelog.d/fixed-cline-install-retry.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Hive and contributor image builds retry Cline installation with fresh npm metadata up to four times to tolerate transient registry propagation failures ([#16666](https://github.com/hivecommons/hive/issues/16666)).
1 change: 1 addition & 0 deletions changelog.d/fixed-merge-celebration-external.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Activity Monitor: a PR merged by anyone — a maintainer's `gh pr merge --admin` as much as the hive's own sweep — now leaves the board with the bee fly-off and a merge toast naming the merger (`#16644 merged by @clubanderson` / `merged by 🐝 hive`); `recently_merged` is the union of the audit log and the merged ledger with `merged_by`, an `issue_comment` on a just-merged PR no longer reads as "closed", and a closed-not-merged exit keeps its quiet toast (hivecommons/hive#16658).
1 change: 1 addition & 0 deletions changelog.d/fixed-pr-progress-chevron-fallback.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Activity Monitor: the PR in progress chevron is the row's fallback — it defers (settled, icon only) whenever CI or Review runs, so CI + Review is the only pair ever active together — and its words are the PR's lifecycle (`Unassigned · next: scanner`, `Waiting for @author · fix owed`), never the CI / review verdict — and an active CI / Review chevron has exactly one fill: the stalled-blocker track no longer renders a darker amber, and running checks stay active whatever the poll's age (hivecommons/hive#16643).
1 change: 1 addition & 0 deletions changelog.d/fixed-rereview-pill-attribution.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Activity Monitor: a pushed head with nothing against it now waits on `🐝 hive reviewer · re-review of <sha>` instead of the PR author (`rework.owed_by`), the Review chevron is the lit one once CI is green, and the GraphQL PR batch no longer caches a truncated `statusCheckRollup` page as the head's full check list — which had read `CI 7/11 · required check "test" hasn't reported` on an all-green head with 70+ contexts (#16635).
1 change: 1 addition & 0 deletions changelog.d/fixed-review-track-never-unknown.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- The Activity Monitor's Review and CI chevrons never say "activity unknown": with no verdict and no reviewer in flight the tooltip states what the head waits on (closed, draft, waiting for CI, not started and who is next, or no data for the head yet · refreshing), a settled verdict names who decided, how long ago and each perspective's verdict with its first finding, and settled verdicts stay known as observations age.
1 change: 1 addition & 0 deletions changelog.d/fixed-sentinel-pacing-selfreport.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Avoid filing repeated-error fleet reports for sentinel skips caused by GitHub pacing, while retaining the local warnings and reporting other failures.
1 change: 1 addition & 0 deletions changelog.d/fixed-signoff-rehome-hive-authored.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Hive no longer supersedes its own PRs for a missing sign-off: the automerge DCO stage treats commits on a PR by hive's `ai_author` or App login as self-squash (so an App-signed commit no longer counts as squash-unsafe), `signoff-rehome` keeps hive-authored and operator-approved heads in place instead of re-authoring them as a new PR, the App's signed re-author commit always carries a `Signed-off-by` trailer, and the push relay refuses a hive-identity commit that lacks one.
1 change: 1 addition & 0 deletions changelog.d/fixed-superseder-signoff-park.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Review aggregation no longer routes a hive-superseding PR to the fixer when every perspective's `requires_human` is a pure approver question with nothing above an informational finding. That route dispatched no-op fix kicks until the cycle was exhausted, then re-authored the same change (#16398 → #16456 → #16607) while the approver the reviewer asked never saw a decision. Such a head keeps its decision question, parks at `stage/085-needs-decision` with approve / request-changes actions, and the tracker's review chevron reads "waiting on a human approver · reviewer verified, needs sign-off".
1 change: 1 addition & 0 deletions changelog.d/security-16638-workflow-review.md
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
- Require human policy sign-off for reviewer questions on workflow and action edits, including same-repository PRs.
11 changes: 10 additions & 1 deletion src/Dockerfile
Original file line number Diff line number Diff line change
Expand Up @@ -319,7 +319,16 @@ RUN timeout 120s npm install -g --ignore-scripts --no-fund --no-audit \
# static list, and non-codex backends are unaffected.
ARG CODEX_VERSION=0.166.0
ARG CLINE_VERSION=3.0.74
RUN npm install -g --ignore-scripts --no-fund --no-audit cline@${CLINE_VERSION} \
# npm can briefly advertise AWS SDK versions before all registry replicas serve
# them. Wait at most three minutes between attempts and revalidate metadata.
RUN attempt=1; \
until npm install -g --ignore-scripts --no-fund --no-audit --prefer-online cline@${CLINE_VERSION}; do \
status=$?; \
if [ "$attempt" -ge 4 ]; then exit "$status"; fi; \
echo "Cline install failed (attempt $attempt/4); retrying in 60s" >&2; \
sleep 60; \
attempt=$((attempt + 1)); \
done \
&& ln -sf "$(npm root -g)/cline/bin/cline" /usr/local/bin/cline \
&& cline --version
RUN npm install -g --ignore-scripts --no-fund --no-audit @openai/codex@${CODEX_VERSION} && npm cache clean --force || \
Expand Down
11 changes: 10 additions & 1 deletion src/Dockerfile.contributor
Original file line number Diff line number Diff line change
Expand Up @@ -243,7 +243,7 @@
# NODESOURCE_KEY_SHA256 is the digest of the raw ASCII-armored key as
# published at the URL below; re-verify against NodeSource's documented
# fingerprint (6F71F5252828 41EEDAF851B4 2F59B5F99B1B E0B4) before bumping.
ARG NODESOURCE_KEY_SHA256=b42e0321dabdc24e892115da705cf061167eac12a317f23d329862d0aa0a271d

Check warning on line 246 in src/Dockerfile.contributor

View workflow job for this annotation

GitHub Actions / build-contributor (linux/arm64, true, true, 75)

Sensitive data should not be used in the ARG or ENV commands

SecretsUsedInArgOrEnv: Do not use ARG or ENV instructions for sensitive data (ARG "NODESOURCE_KEY_SHA256") More info: https://docs.docker.com/go/dockerfile/rule/secrets-used-in-arg-or-env/

Check warning on line 246 in src/Dockerfile.contributor

View workflow job for this annotation

GitHub Actions / build-contributor (linux/amd64, true, false, 60)

Sensitive data should not be used in the ARG or ENV commands

SecretsUsedInArgOrEnv: Do not use ARG or ENV instructions for sensitive data (ARG "NODESOURCE_KEY_SHA256") More info: https://docs.docker.com/go/dockerfile/rule/secrets-used-in-arg-or-env/
RUN mkdir -p /etc/apt/keyrings \
&& curl -fsSL -o /tmp/nodesource.key https://deb.nodesource.com/gpgkey/nodesource-repo.gpg.key \
&& echo "${NODESOURCE_KEY_SHA256} /tmp/nodesource.key" | sha256sum -c - \
Expand Down Expand Up @@ -575,7 +575,16 @@
# proactive quota protection with no custom reader and no local hive server.
COPY --from=quota-publisher-builder /hive-quota-publisher /usr/local/bin/hive-quota-publisher
COPY --from=quota-publisher-builder /crux /usr/local/bin/crux
RUN npm install -g --ignore-scripts --no-fund --no-audit cline@${CLINE_VERSION} \
# npm can briefly advertise AWS SDK versions before all registry replicas serve
# them. Wait at most three minutes between attempts and revalidate metadata.
RUN attempt=1; \
until npm install -g --ignore-scripts --no-fund --no-audit --prefer-online cline@${CLINE_VERSION}; do \
status=$?; \
if [ "$attempt" -ge 4 ]; then exit "$status"; fi; \
echo "Cline install failed (attempt $attempt/4); retrying in 60s" >&2; \
sleep 60; \
attempt=$((attempt + 1)); \
done \
&& ln -sf "$(npm root -g)/cline/bin/cline" /usr/local/bin/cline \
&& cline --version
COPY config/backends.conf /usr/local/etc/hive/
Expand Down
18 changes: 18 additions & 0 deletions src/cmd/crux/main_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -44,3 +44,21 @@ func TestRunReportsVersion(t *testing.T) {
t.Fatalf("code %d, stdout %q", code, stdout.String())
}
}

// A bad OTEL configuration costs telemetry, never the command
// (hivecommons/crux#19): run reports the refusal on stderr and the verb still
// completes with its own exit code.
func TestRunKeepsCommandWhenTelemetryConfigIsBad(t *testing.T) {
t.Setenv("OTEL_EXPORTER_OTLP_ENDPOINT", "http://127.0.0.1:1")
t.Setenv("OTEL_EXPORTER_OTLP_PROTOCOL", "grpc")
var stdout, stderr bytes.Buffer
code := run([]string{"--version"}, strings.NewReader(""), &stdout, &stderr, func() (context.Context, context.CancelFunc) {
return context.WithCancel(context.Background())
})
if code != cli.ExitOK || stdout.String() != "crux "+cli.DefaultVersion+"\n" {
t.Fatalf("code %d, stdout %q, stderr %q", code, stdout.String(), stderr.String())
}
if !strings.Contains(stderr.String(), "crux: telemetry disabled:") || !strings.Contains(stderr.String(), "OTEL_EXPORTER_OTLP_PROTOCOL") {
t.Fatalf("stderr = %q, want the telemetry-disabled notice naming the bad variable", stderr.String())
}
}
1 change: 1 addition & 0 deletions src/cmd/hive/merge_eligibility.go
Original file line number Diff line number Diff line change
Expand Up @@ -146,6 +146,7 @@ func stageLabelPolicyFunc(cfg *config.Config) func() automerge.StageLabelPolicy
AuthorKindLabels: cfg.Governor.Labels.Author.Names(),
RelayKnown: contributorRelayKnown,
OwnerAuthor: dashboardOwnerLogin(cfg),
HiveAuthor: hiveIdentity(cfg).Matches,
SharedCIIncident: stageSharedCIIncident,
BaseBranchRed: stageBaseBranchRed,
ForkGuard: stageForkGuard,
Expand Down
16 changes: 16 additions & 0 deletions src/crux/cmd/crux/main_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -50,6 +50,22 @@ func TestRunReportsVersion(t *testing.T) {
}
}

// A bad OTEL configuration costs telemetry, never the command
// (hivecommons/crux#19): run reports the refusal on stderr and the verb still
// completes with its own exit code.
func TestRunKeepsCommandWhenTelemetryConfigIsBad(t *testing.T) {
t.Setenv("OTEL_EXPORTER_OTLP_ENDPOINT", "http://127.0.0.1:1")
t.Setenv("OTEL_EXPORTER_OTLP_PROTOCOL", "grpc")
var stdout, stderr bytes.Buffer
code := run([]string{"--version"}, strings.NewReader(""), &stdout, &stderr, testCtx)
if code != cli.ExitOK || stdout.String() != "crux "+cli.DefaultVersion+"\n" {
t.Fatalf("code %d, stdout %q, stderr %q", code, stdout.String(), stderr.String())
}
if !strings.Contains(stderr.String(), "crux: telemetry disabled:") || !strings.Contains(stderr.String(), "OTEL_EXPORTER_OTLP_PROTOCOL") {
t.Fatalf("stderr = %q, want the telemetry-disabled notice naming the bad variable", stderr.String())
}
}

func TestRunPSFromDefsDirectory(t *testing.T) {
dir := t.TempDir()
defsDir := filepath.Join(dir, "defs")
Expand Down
90 changes: 90 additions & 0 deletions src/deploy/test_cline_install_retry.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,90 @@
#!/usr/bin/env python3
"""Exercise the shipped Cline RUN commands without npm, network, or Docker."""

import os
from pathlib import Path
import re
import subprocess
import tempfile
import unittest


SRC = Path(__file__).resolve().parents[1]
MOCKS = r'''
count=0
npm() {
if [ "$1" = root ]; then echo /mock/node_modules; return 0; fi
printf 'npm %s\n' "$*" >> "$CALL_LOG"
count=$((count + 1))
if [ "$count" -le "$FAILURES" ]; then
echo 'npm error code ETARGET: No matching version found' >&2
return 42
fi
}
sleep() { echo "sleep $*" >> "$CALL_LOG"; }
ln() { echo "ln $*" >> "$CALL_LOG"; }
cline() { echo "cline $*" >> "$CALL_LOG"; return "$VERSION_STATUS"; }
'''


class ClineInstallRetryTest(unittest.TestCase):
def run_install(self, dockerfile, failures, version_status=0):
# Join Docker continuations, then run the actual RUN body under /bin/sh
# just like the images. No copied retry implementation in this test.
text = (SRC / dockerfile).read_text().replace("\\\n", " ")
commands = [
line.removeprefix("RUN ")
for line in text.splitlines()
if line.startswith("RUN ")
and re.search(r"npm install .*cline@", line)
]
self.assertEqual(len(commands), 1, dockerfile)
with tempfile.TemporaryDirectory() as directory:
log = Path(directory) / "calls"
log.touch()
result = subprocess.run(
["/bin/sh", "-c", MOCKS + commands[0]],
env={**os.environ, "CALL_LOG": str(log),
"FAILURES": str(failures), "CLINE_VERSION": "3.0.74",
"VERSION_STATUS": str(version_status)},
capture_output=True, text=True, timeout=5,
)
calls = log.read_text().splitlines()
return result, calls

def test_install_retry_and_failure(self):
for dockerfile in ("Dockerfile", "Dockerfile.contributor"):
# Includes success on the final allowed attempt and exhaustion.
for failures in (0, 1, 3, 4):
with self.subTest(dockerfile=dockerfile, failures=failures):
result, calls = self.run_install(dockerfile, failures)
self.assertEqual(result.returncode, 42 if failures == 4 else 0,
result.stderr)
installs = [call for call in calls if call.startswith("npm ")]
self.assertEqual(len(installs), min(failures + 1, 4))
for install in installs:
self.assertEqual(install.split(), [
"npm", "install", "-g", "--ignore-scripts", "--no-fund",
"--no-audit", "--prefer-online", "cline@3.0.74",
])
self.assertEqual(calls.count("sleep 60"), min(failures, 3))
self.assertEqual([call for call in calls if call.startswith("sleep ")],
["sleep 60"] * min(failures, 3))
post_install = [call for call in calls
if call.startswith(("ln ", "cline "))]
self.assertEqual(post_install, [] if failures == 4 else [
"ln -sf /mock/node_modules/cline/bin/cline /usr/local/bin/cline",
"cline --version",
])

def test_version_failure_is_not_retried_or_masked(self):
for dockerfile in ("Dockerfile", "Dockerfile.contributor"):
with self.subTest(dockerfile=dockerfile):
result, calls = self.run_install(dockerfile, 0, version_status=17)
self.assertEqual(result.returncode, 17, result.stderr)
self.assertEqual(sum(call.startswith("npm ") for call in calls), 1)
self.assertFalse(any(call.startswith("sleep ") for call in calls))


if __name__ == "__main__":
unittest.main()
8 changes: 8 additions & 0 deletions src/deploy/test_supply_chain_pins.sh
Original file line number Diff line number Diff line change
Expand Up @@ -269,6 +269,14 @@ else
fi
fi

# Execute the real Cline install RUN bodies with mocked npm/links/sleep. This
# keeps the bounded retry covered by the existing supply-chain CI step.
if python3 "$V2_DIR/deploy/test_cline_install_retry.py"; then
pass "Cline installs tolerate transient registry failures and fail closed"
else
fail "Cline install retry behavior"
fi

echo
echo "=== $PASS passed, $FAIL failed ==="
[ "$FAIL" -eq 0 ] || exit 1
21 changes: 21 additions & 0 deletions src/docs/acmm-policy-matrix.md
Original file line number Diff line number Diff line change
Expand Up @@ -16,6 +16,27 @@ issues. The **KB** is the knowledge base of reusable facts; an **RFC** is a
request for comments on a proposed design. For a first deployment's trust
choices, start with [Securing your hive](securing-your-hive.md).

## Repository E2E detection

The codebase evaluator's `acmm:prereq-e2e` prerequisite accepts the existing
Playwright/Cypress configuration and E2E directory conventions, as well as
GitHub Actions workflow files named `test*-install*.yml`, `test*-boot*.yml`,
or `*e2e*.yml` (also `.yaml`). These workflow alternatives cover projects
whose end-to-end boundary is installing and booting a system rather than
driving a web browser.

Workflow patterns use Go `path.Match` semantics on the file basename, only
inside the literal `.github/workflows` directory. They do not recurse, match
directories named like workflow files, or accept an unrelated CI workflow
just because it exists. Detection remains a file-existence heuristic, not
proof that a workflow passes or that its contents implement an E2E test;
maintainers must keep these paths backed by real coverage rather than add
empty directories or dummy workflows to satisfy the criterion.

An evaluator fix takes effect only after a release containing it is deployed.
After deployment, refresh the ACMM evaluation (subject to its refresh debounce)
before closing an existing false-negative gap issue.

## Policy Modes

The four policy tiers below describe the main autonomy ladder. The roster also
Expand Down
Loading
Loading