Compare commits

..
Author SHA1 Message Date
Jason WoltjeandClaude Fable 5 74b3b9e869 fix(git-tools): ci-queue-wait mishandles Gitea statuses:null — push guard hard-fails on any repo without CI
ci/woodpecker/pr/ci Pipeline failed
Gitea's combined-status endpoint returns statuses:null (not []) plus a
synthetic aggregate state of "pending" for a commit with zero status
contexts (captured live 2026-08-09). The bash parser called that payload
malformed -> ASSERTED_NOT_READY exit 3, blocking every push to a CI-less
repo; hit twice by velma on two independent clones (shared checkout +
fresh sparse clone), proving it environment-independent.

- treat statuses:null as empty (both .sh and .ps1)
- classify zero contexts as no-status BEFORE consulting the synthetic
  aggregate state (both twins; honoring it would poll to timeout)
- no-status on --purpose push without --require-status is now queue-clear
  exit 0 (a repo with no CI has no queue), mirroring record_cannot_assert
  dispositions (push=degraded-pass, merge=hold); merge + --require-status
  stay fail-closed at exit 3
- red-first regression harness test-ci-queue-wait-no-status.sh (5 cases,
  incl. the verbatim live Gitea payload and a genuine-pending guard);
  branch-absent harness still green

Co-Authored-By: Claude Fable 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_013SAYFkRhQfhguY7AHfiUC8
2026-08-09 03:02:25 -05:00
5 changed files with 218 additions and 1832 deletions
File diff suppressed because it is too large Load Diff
-114
View File
@@ -1,114 +0,0 @@
# WebUI Fleet Bridge Planning Scratchpad
**Mode:** Task 0 docs-only decision PR authorized; Task 1 and all executable/runtime work remain blocked.
**Owner:** Velma
**Opened:** 2026-08-09
**Scope:** One enrolled Agent Host launching one native `mosaic claude` OAuth session and streaming it into `apps/web` through `apps/gateway`.
## Objective
Turn the approved direction in `jarvis-brain/docs/scratchpads/MOSAIC-WEBUI-FLEET-BRIDGE.md` into a test-first implementation plan while preserving Fred's harness-home/launcher contract and Scooby's greenfield safety findings.
## Source reconciliation
- Current `origin/main`: `b0f7d26dd9c14d91eaaefc35d6c9fd6618a0bd92`.
- Current `origin/next`: `4df478cdd150fdf8d52ea109f02ade5d85017acd`.
- Branches currently diverge (`main` has 11 unique commits; `next` has 13). `next` contains local-tier Redis fix #689; `main` contains later fleet/shell fixes.
- Fred's three-root harness-home design and promotion stack are not yet fully present on either baseline.
- Therefore neither current SHA is an acceptable implementation pin. Code may begin only from a Fred-certified reconciled SHA containing the required launcher/home contract and safe Gateway startup prerequisites.
- The deployed `mosaic.woltje.com` v0.0.20 UI remains reference-only.
## Verified seams
- Current browser chat uses an in-process Pi SDK session.
- `AgentRuntimeProvider` supports list/tree/stream/send/attach/detach/terminate, but not create/start.
- `InteractionController` enrolls an already-existing runtime session; it cannot launch one.
- Hermes is the only runtime provider registered in Gateway.
- Tmux streaming is explicitly unsupported and remains out of scope.
- `mosaic claude` is the authoritative launcher and accepts Claude's machine-facing stream-json flags.
- Installed discovery version: Claude Code 2.1.226. Target Distrobox version must be independently pinned and certified.
## Non-negotiable dependencies
1. Fred approves the machine-facing launcher/seat-home contract before code.
2. No WebUI/Gateway direct read of lease broker state, daemon socket, or state files (F-V3).
3. No provider OAuth token leaves the Agent Host.
4. No local Gateway/Web startup around the KBN/database hold.
5. Greenfield work runs in a Debian Distrobox with an isolated home.
6. The initial plan PR targets `next`; Fred binds D1D15 on its exact head before issue/PRD/tracker completion or implementation work.
## Reproducible evidence
Run from a clean Stack clone:
```bash
git fetch origin main next
git rev-parse origin/main origin/next
git rev-list --left-right --count origin/main...origin/next
rg -n "interface AgentRuntimeProvider|createSession|streamEvents|terminate" \
packages/types/src/agent packages/agent/src apps/gateway/src/agent
rg -n "AgentService\.prompt|interaction_sessions|createRuntimeTerminationApproval" \
apps/gateway/src packages/db/src/schema.ts
```
Primary inspected source seams:
- `packages/types/src/agent/agent-runtime-provider.ts`
- `packages/agent/src/{runtime-provider-registry,hermes-runtime-provider,matrix-native-runtime-provider,tmux-fleet-runtime-provider}.ts`
- `apps/gateway/src/agent/{runtime-provider-registry.service,interaction.controller,durable-session.repository,durable-session.service}.ts`
- `apps/gateway/src/chat/chat.gateway.ts`
- `packages/mosaic/src/commands/{launch,interaction}.ts`
- `packages/mosaic/src/fleet/generated-env-boundary.ts`
- `packages/db/src/schema.ts`
- `apps/web/src/app/(dashboard)/chat/page.tsx`
Planning-only investigation transcripts are local and intentionally uncommitted:
- `/tmp/velma-plan-stack-surface.txt`
- `/tmp/velma-plan-structure.txt`
- `/tmp/velma-plan-scooby.txt`
- `/tmp/velma-plan-runtime-contract.txt`
## Source findings that constrain the design
- `interaction_sessions.id` is the stable primary key; there are no create/policy/enrollment/state columns.
- `interaction_outbox` has a unique `(session_id, idempotency_key)` index and only `pending | processing | delivered`.
- Baseline `DurableSessionRepository.create()` can replace provider/runtime identity for the same owner; M1 must remove that implicit mutation.
- Baseline termination approval is Redis-backed and currently consumes separately from PostgreSQL; M1 therefore needs durable authorization acceptance before destructive token deletion/dispatch.
- Existing interaction HTTP base is `/api/interaction/:agentName`; the plan extends it rather than inventing a second route family.
- Baseline `launch.ts`/lease launcher still use ambient lookup/literal interpreters. Section 4.1 is non-binding consumer input to Fred's W-F design; W-F's final resolved-launch contract must exist in the certified base before Velma can certify it.
- Root `pnpm test` is not KBN-safe: it includes PGlite migration and framework-shell/lease-broker suites.
## Independent draft review
Seven adversarial review rounds found and drove explicit fixes for:
- a candidate resolved-seat consumer descriptor and threat model, now explicitly non-binding input to Fred's W-F-owned launcher design;
- one active launch per stable conversation, exact pending/failed encodings, durable pre-dispatch reservation, CAS activation, and crash lookup without a migration;
- generation-bound enrollment, command/event revalidation, `SIGHUP` config epochs, and stale-epoch rejection;
- Redis/PostgreSQL exact-stop crash safety via non-destructive verify, durable authorization acceptance, atomic claim/`GETDEL`, and same-operation status reconciliation;
- sequence gaps/reorder limits, deterministic UUIDv5 completion, and Gateway restart fail-closed behavior;
- shared streaming redaction before host ring/transport and again before Gateway persistence/browser;
- D2/D12-selected path-free provenance and a separate safe browser DTO—never raw or hashed path strings;
- exact F-V3 boundary: bridge has no broker API, while Fred's sealed launcher may enforce broker policy internally;
- migration-free focused tests in Tasks 18, with live OAuth, repository transaction, Gateway/Web, and Playwright restricted to the Fred/Scooby-certified Task 9 path;
- candidate private/public commitment and artifact-binding mechanisms that W-F may accept, simplify, replace, or defer;
- JCS event-digest recomputation and equal/different duplicate handling in both accepted and future-buffer states;
- a safe browser presentation DTO for host/workspace/seat/persona labels, readiness, and connection state;
- explicit `OnApplicationBootstrap` create/stop recovery enumeration with no auto-launch;
- all POSIX/Windows/UNC/file/tilde path classes in the streaming redactor and definitive failed-stop response semantics;
- the full 14-column canonical task schema, fake-only Task 5 repository tests, migration-free certified Task 9 DB test, server-owned operation correlations/routes, and exact merged-SHA smoke;
- per-commit independent review, queue guards, exact-head PR review, squash merge, exact merged-next SHA/CI wait, worktree-bound smoke, issue-state readback, and reviewed tracker-closure PR with its own merged-next CI;
- a private DB URL loader captured/exported per session without recording or echoing the credential;
- an attached Task 9 implementation branch, provider-filtered exact issue-state readback, and machine-verified smoke JSON binding source/worktree/deployed SHA;
- a capped, duplicate-key-rejecting, exact-key/type smoke schema so report extensions or JSON boolean/float coercion cannot smuggle data or fake child-count evidence.
The plan remains deliberately **decision-PR-ready, not implementation-ready**. Fred authorized only Task 0's initial two-document PR on `next`. He must still return every Section 2 value, replace all `[FRED-GATE]` entries, certify storage/startup, and provide the W-F-dependent `IMPLEMENTATION_BASE_SHA` before Task 0 closes or Task 1 starts.
## Current status
- Gitea principal verified as `velma`; helper and API wrapper resolution are fail-closed and correct.
- Fred authorized the initial Task 0 docs-only PR in `comms/20260809T094952Z__from-fred__ec0e85.md` and confirmed its `next` lane/W-F descriptor corrections in `comms/20260809T095437Z__from-fred__4ee79d.md`.
- Provisional decisions: D1/D8/D9/D10/D13/D14 approved; D4 tool labels exactly `{Read, Grep, Glob}`; D5/D12 provisional; D11 remains a single-operator seam; D2/D3/D6/D7 and `IMPLEMENTATION_BASE_SHA` are `BLOCKED-ON-W-F`.
- Awaiting exact plan-PR-head review and Fred's complete D1D15 binding contract.
- No source code, database, deployment, or live runtime changes made.
@@ -85,10 +85,6 @@ function Get-QueueState {
$state = "$($Payload.state)".ToLowerInvariant() $state = "$($Payload.state)".ToLowerInvariant()
} }
if ($pending -contains $state) { return "pending" }
if ($failure -contains $state) { return "terminal-failure" }
if ($success -contains $state) { return "terminal-success" }
$values = @() $values = @()
$statuses = @() $statuses = @()
if ($null -ne $Payload.statuses) { $statuses = @($Payload.statuses) } if ($null -ne $Payload.statuses) { $statuses = @($Payload.statuses) }
@@ -101,7 +97,15 @@ function Get-QueueState {
if (-not [string]::IsNullOrEmpty($v)) { $values += $v } if (-not [string]::IsNullOrEmpty($v)) { $values += $v }
} }
if ($values.Count -eq 0 -and [string]::IsNullOrEmpty($state)) { return "no-status" } # Zero contexts is classified FIRST: Gitea reports a synthetic aggregate
# state of "pending" alongside statuses:null / total_count:0 (a commit
# with no CI at all), and honoring that aggregate would poll to the
# timeout. With zero contexts there is nothing to wait on.
if ($values.Count -eq 0) { return "no-status" }
if ($pending -contains $state) { return "pending" }
if ($failure -contains $state) { return "terminal-failure" }
if ($success -contains $state) { return "terminal-success" }
if (($values | Where-Object { $pending -contains $_ }).Count -gt 0) { return "pending" } if (($values | Where-Object { $pending -contains $_ }).Count -gt 0) { return "pending" }
if (($values | Where-Object { $failure -contains $_ }).Count -gt 0) { return "terminal-failure" } if (($values | Where-Object { $failure -contains $_ }).Count -gt 0) { return "terminal-failure" }
if ($values.Count -gt 0 -and ($values | Where-Object { -not ($success -contains $_) }).Count -eq 0) { return "terminal-success" } if ($values.Count -gt 0 -and ($values | Where-Object { -not ($success -contains $_) }).Count -eq 0) { return "terminal-success" }
@@ -52,7 +52,11 @@ except Exception:
print("malformed") print("malformed")
raise SystemExit(0) raise SystemExit(0)
# Gitea returns "statuses": null (not []) for a commit with zero status
# contexts -- e.g. any repo with no CI configured. Treat null as empty.
raw_statuses = payload.get("statuses", []) raw_statuses = payload.get("statuses", [])
if raw_statuses is None:
raw_statuses = []
raw_state = payload.get("state", "") raw_state = payload.get("state", "")
if not isinstance(raw_statuses, list) or not isinstance(raw_state, str): if not isinstance(raw_statuses, list) or not isinstance(raw_state, str):
print("malformed") print("malformed")
@@ -75,14 +79,18 @@ for item in statuses:
raise SystemExit(0) raise SystemExit(0)
values.append(raw_value.lower()) values.append(raw_value.lower())
if any(value in pending_values for value in values) or state in pending_values: # Zero contexts is classified FIRST: Gitea reports a synthetic aggregate
# state of "pending" alongside total_count:0, and an aggregate with no
# contexts behind it must not read as an in-flight pipeline (it would poll
# to the timeout). With zero contexts there is nothing to wait on.
if not values:
print("no-status")
elif any(value in pending_values for value in values) or state in pending_values:
print("pending") print("pending")
elif any(value in failure_values for value in values) or state in failure_values: elif any(value in failure_values for value in values) or state in failure_values:
print("terminal-failure") print("terminal-failure")
elif values and all(value in success_values for value in values) and state in {"", "success"}: elif all(value in success_values for value in values) and state in {"", "success"}:
print("terminal-success") print("terminal-success")
elif not values:
print("no-status")
else: else:
print("unknown") print("unknown")
' '
@@ -465,9 +473,18 @@ while true; do
no-status) no-status)
if [[ "$REQUIRE_STATUS" -eq 1 ]]; then if [[ "$REQUIRE_STATUS" -eq 1 ]]; then
echo "Error: ASSERTED_NOT_READY state=no-status; --require-status was set for ${BRANCH}." >&2 echo "Error: ASSERTED_NOT_READY state=no-status; --require-status was set for ${BRANCH}." >&2
else exit 3
echo "Error: ASSERTED_NOT_READY state=no-status purpose=${PURPOSE} branch=${BRANCH}." >&2
fi fi
# A head with zero status contexts has no CI queue to wait on.
# For push, that is queue-clear (a repo with no CI must remain
# pushable) -- mirroring record_cannot_assert's dispositions
# (push=degraded-pass, merge=hold). Merge stays fail-closed:
# no-status there may just mean CI has not reported yet.
if [[ "$PURPOSE" == "push" ]]; then
echo "[ci-queue-wait] no status contexts on ${BRANCH} head (state=no-status) — no CI queue to wait on; queue clear."
exit 0
fi
echo "Error: ASSERTED_NOT_READY state=no-status purpose=${PURPOSE} branch=${BRANCH}." >&2
exit 3 exit 3
;; ;;
terminal-failure|malformed|unknown) terminal-failure|malformed|unknown)
@@ -0,0 +1,186 @@
#!/usr/bin/env bash
# Regression harness for ci-queue-wait.sh's zero-status-context handling.
#
# Gitea's combined-status endpoint returns, for a commit with NO status
# contexts (e.g. a repo with no CI configured at all):
# {"state":"pending","sha":"...","total_count":0,"statuses":null,...}
# -- statuses is JSON null (not []), and state is a synthetic "pending"
# even though nothing is running. Captured live from a Gitea 1.22 host,
# 2026-08-09 (jason.woltje/jarvis-brain, a repo with no pipeline).
#
# Before the fix, `payload.get("statuses", [])` received null, failed the
# isinstance(list) check, and the guard reported ASSERTED_NOT_READY
# state=malformed (exit 3) -- blocking every push to a CI-less repo. Had
# null been tolerated, the synthetic aggregate "pending" would instead
# have polled to the timeout (exit 124). The fix must:
# 1. treat statuses:null as an empty list, and
# 2. classify zero status VALUES as "no-status" regardless of the
# synthetic aggregate state, and
# 3. treat no-status for --purpose push (without --require-status) as
# queue-clear (exit 0) -- a repo with no CI has no queue to wait on;
# this mirrors record_cannot_assert's disposition table
# (push=degraded-pass, merge=hold).
#
# Covers:
# (a) statuses:null + synthetic state:pending, purpose=push
# -> exit 0, queue-clear/no-status message (THE live fault).
# (b) same payload, purpose=merge -> still fail-closed (exit 3).
# (c) same payload, push + --require-status -> still fail-closed (exit 3).
# (d) statuses:[] + state:"" -> same as (a) (exit 0).
# (e) a real pending context -> still polls (times out, 124),
# proving the relaxation didn't swallow genuine pending states.
set -euo pipefail
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
WORK_DIR="${MOSAIC_TEST_WORK_DIR:-$PWD/.mosaic-test-work/ci-queue-wait-no-status}"
REPO_DIR="$WORK_DIR/repo"
STUB_DIR="$WORK_DIR/stubs"
rm -rf "$WORK_DIR"
mkdir -p "$REPO_DIR" "$STUB_DIR"
git -C "$REPO_DIR" init -q
git -C "$REPO_DIR" remote add origin https://git.example.test/acme/widgets.git
# Minimal curl stub (same conventions as test-ci-queue-wait-branch-absent.sh):
# branch lookup answers 200 with a fixed SHA; the status endpoint's payload
# is selected by MOSAIC_STUB_STATUS_MODE.
cat > "$STUB_DIR/curl" <<'SH'
#!/usr/bin/env bash
set -euo pipefail
has_w=0
url=""
for arg in "$@"; do
case "$arg" in
-w) has_w=1 ;;
http://*|https://*) url="$arg" ;;
esac
done
case "$url" in
*/branches/*)
body='{"commit":{"id":"deadbeefcafef00d0123456789abcdef01234567"}}'
if [[ "$has_w" == 1 ]]; then
printf '%s\n200' "$body"
else
printf '%s' "$body"
fi
exit 0
;;
*/status)
mode="${MOSAIC_STUB_STATUS_MODE:?MOSAIC_STUB_STATUS_MODE not set}"
;;
*)
echo "curl stub: unrecognized URL: $url" >&2
exit 2
;;
esac
case "$mode" in
# Verbatim shape of the live Gitea response (repository object elided).
gitea-null-statuses)
body='{"state":"pending","sha":"deadbeefcafef00d0123456789abcdef01234567","total_count":0,"statuses":null,"url":"","commit_url":""}'
;;
empty-statuses)
body='{"state":"","statuses":[]}'
;;
real-pending)
body='{"state":"pending","statuses":[{"context":"ci/woodpecker","status":"running","target_url":""}]}'
;;
*)
echo "curl stub: unknown mode=$mode" >&2
exit 2
;;
esac
printf '%s' "$body"
SH
chmod +x "$STUB_DIR/curl"
run_ci_queue_wait() {
(
cd "$REPO_DIR"
export PATH="$STUB_DIR:$PATH"
export MOSAIC_CREDENTIALS_FILE="$WORK_DIR/no-credentials.json"
export MOSAIC_CI_QUEUE_AUDIT_LOG="$WORK_DIR/audit.jsonl"
export GITEA_TOKEN="stub-token"
export GITEA_URL="https://git.example.test"
"$SCRIPT_DIR/ci-queue-wait.sh" -B main -t 3 -i 1 "$@"
)
}
fail=0
# (a) THE live fault: statuses:null + synthetic pending, purpose=push -> exit 0.
set +e
out_a=$(MOSAIC_STUB_STATUS_MODE=gitea-null-statuses run_ci_queue_wait --purpose push 2>&1)
status_a=$?
set -e
if [[ "$status_a" -ne 0 ]]; then
echo "FAIL(a): expected exit 0 for null-statuses/no-CI repo on push, got $status_a" >&2
echo "$out_a" >&2
fail=1
elif [[ "$out_a" == *"malformed"* ]]; then
echo "FAIL(a): null statuses must not be classified as malformed" >&2
echo "$out_a" >&2
fail=1
elif [[ "$out_a" != *"no-status"* ]]; then
echo "FAIL(a): expected a no-status/queue-clear message, got:" >&2
echo "$out_a" >&2
fail=1
fi
# (b) Same payload, purpose=merge -> hold, fail-closed exit 3.
set +e
out_b=$(MOSAIC_STUB_STATUS_MODE=gitea-null-statuses run_ci_queue_wait --purpose merge 2>&1)
status_b=$?
set -e
if [[ "$status_b" -ne 3 ]]; then
echo "FAIL(b): expected exit 3 for no-status on merge, got $status_b" >&2
echo "$out_b" >&2
fail=1
fi
# (c) Same payload, push + --require-status -> strictness opt-in still fails.
set +e
out_c=$(MOSAIC_STUB_STATUS_MODE=gitea-null-statuses run_ci_queue_wait --purpose push --require-status 2>&1)
status_c=$?
set -e
if [[ "$status_c" -ne 3 ]]; then
echo "FAIL(c): expected exit 3 for --require-status with no contexts, got $status_c" >&2
echo "$out_c" >&2
fail=1
fi
# (d) statuses:[] + state:"" (the shape the code always tolerated) -> exit 0 on push.
set +e
out_d=$(MOSAIC_STUB_STATUS_MODE=empty-statuses run_ci_queue_wait --purpose push 2>&1)
status_d=$?
set -e
if [[ "$status_d" -ne 0 ]]; then
echo "FAIL(d): expected exit 0 for empty-statuses on push, got $status_d" >&2
echo "$out_d" >&2
fail=1
fi
# (e) A REAL pending context must still block: polls to timeout, exit 124.
set +e
out_e=$(MOSAIC_STUB_STATUS_MODE=real-pending run_ci_queue_wait --purpose push 2>&1)
status_e=$?
set -e
if [[ "$status_e" -ne 124 ]]; then
echo "FAIL(e): expected exit 124 (timeout) for a genuinely pending context, got $status_e" >&2
echo "$out_e" >&2
fail=1
elif [[ "$out_e" != *"ci/woodpecker=running"* ]]; then
echo "FAIL(e): expected the pending context to be reported, got:" >&2
echo "$out_e" >&2
fail=1
fi
if [[ "$fail" -eq 0 ]]; then
echo "ci-queue-wait no-status regression passed (5/5 cases)"
fi
exit "$fail"