test(synth-e2e): add Files API config.yaml round-trip gate

Today's user-visible bug ("PUT /workspaces/<id>/files/config.yaml: 500
… install: cannot create directory '/opt/configs': Permission denied",
fixed in #2769) shipped to production and was caught only when an
operator opened the Canvas Config tab and clicked Save & Restart on
a claude-code workspace. Two compounding root causes:

1. Path-map fall-through: claude-code wasn't in
   workspaceFilePathPrefix, so it fell through to the /opt/configs
   default — a path the workspace EC2 doesn't have (cloud-init only
   creates /configs).
2. Permission: /configs is root-owned, but the SSH-as-ubuntu install
   command had no sudo prefix, so the write would have failed with
   EACCES even with the right path.

The synth E2E provisions a fresh workspace every cron firing but
never PUTs a file via the Files API. So neither failure mode could
fail the canary.

Add a new step 7c (between terminal-diagnose and A2A) that:
  - PUTs a known marker into config.yaml on each provisioned workspace
  - GETs it back and asserts the marker is present
  - Fails with an actionable message that names the likely class of
    regression (path map vs permission) so the next operator doesn't
    have to re-discover today's debugging path

The marker includes the run ID so stale state from a prior canary
can't false-pass.

Why round-trip (not just PUT-and-200): a 200 from PUT only proves the
SSH install succeeded somewhere on disk; the GET-back proves the file
landed at the path the runtime actually reads from (i.e., that the
host:/configs → container:/configs bind-mount sees it). Without the
GET, a future bug that writes to a non-bind-mounted host path would
silently no-op from the runtime's POV but pass the gate.

Deferred (separate PR, requires AWS-creds wiring): a parallel gate
that aws ec2 describe-instances on the workspace EC2 and asserts the
attached IamInstanceProfile.Arn — would directly catch the #466 IAM
profile gap class. Punted because it needs aws-actions/configure-aws-
credentials added to continuous-synth-e2e.yml + a read-only IAM role
provisioned on the AWS side. Tracked as task #301.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Hongming Wang 2026-05-04 14:43:58 -07:00
parent d866d3aa5f
commit a242ca8b01

View File

@ -504,6 +504,54 @@ for wid in $WS_TO_CHECK; do
fi
done
# ─── 7c. Workspace files API config.yaml round-trip ────────────────────
# Pin the config-save path that drives the Canvas Config tab's Save &
# Restart. Two failure classes this gate catches in one shot:
#
# 1. Path map drift (PR #2769). Runtime falls through to the wrong
# base path (e.g. /opt/configs when user-data only created /configs)
# → SSH `install -D` fails with EACCES on a parent dir that doesn't
# exist. The user-visible 500 was unobservable without exercising
# this code path on a fresh workspace.
# 2. Permission drift on /configs. The path is root-owned by cloud-init,
# so the SSH-as-ubuntu install needs `sudo -n`. Any future change
# that drops the sudo, switches to a non-passwordless-sudo OS user,
# or moves the path to a non-ubuntu-writable dir without sudo will
# regress this gate.
#
# Round-trip: PUT a known marker, GET it back, assert content matches.
# Marker shape includes the run id so a stale file from a prior canary
# can't false-pass.
log "7c/11 Files API config.yaml round-trip..."
CONFIG_MARKER="# molecule-synth-e2e: ${E2E_RUN_ID:-unknown} ${RUNTIME} $(date -u +%Y-%m-%dT%H:%M:%SZ)"
CONFIG_PAYLOAD="${CONFIG_MARKER}
name: synth-canary
runtime: ${RUNTIME}
"
for wid in $WS_TO_CHECK; do
PUT_BODY=$(python3 -c "import json,sys; print(json.dumps({'content': sys.stdin.read()}))" <<< "$CONFIG_PAYLOAD")
PUT_RESP=$(tenant_call PUT "/workspaces/$wid/files/config.yaml" \
-H "Content-Type: application/json" \
-d "$PUT_BODY" \
-w $'\n%{http_code}\n' \
2>/dev/null || printf '\n500\n')
PUT_CODE=$(echo "$PUT_RESP" | tail -n 2 | head -n 1)
PUT_BODY_OUT=$(echo "$PUT_RESP" | sed '$d' | sed '$d')
if [ "$PUT_CODE" != "200" ] && [ "$PUT_CODE" != "204" ]; then
fail "Workspace $wid Files API PUT config.yaml returned $PUT_CODE: $PUT_BODY_OUT — likely a path-map or permission regression in workspace-server template_files_eic.go"
fi
# GET back and assert the marker line is present. Don't require exact
# equality — the runtime's loader may normalize trailing newline /
# quoting; presence of the marker proves the content landed at the
# path the runtime reads from (vs landing at a host path that's
# invisible to the bind-mounted container).
GET_RESP=$(tenant_call GET "/workspaces/$wid/files/config.yaml" 2>/dev/null || echo "")
if ! echo "$GET_RESP" | grep -qF "$CONFIG_MARKER"; then
fail "Workspace $wid Files API GET config.yaml does not contain the marker just written ('$CONFIG_MARKER'). Either the PUT landed at a host path the container doesn't bind-mount, or the GET reads from a different path. Either way, Canvas Save & Restart will appear to succeed but the workspace won't pick up the change."
fi
ok " $wid config.yaml round-trip OK"
done
# ─── 8. A2A round-trip on parent ───────────────────────────────────────
log "8/11 Sending A2A message to parent — expecting agent response..."
# Smoke prompt phrasing — DO NOT trim back to the bare "Reply with exactly: PONG"