Gate image builds on socktop CLI compatibility; fix agent probe; caps for entrypoint

- CI now builds the image locally on the (arm64) runner and runs
  scripts/verify-image-socktop-flags.sh before pushing: every --flag the
  restricted/session shells pass must be documented by the socktop
  binary actually installed in the image. Catches the 0.3.9 failure
  class (cached apt layer shipping a pre-flag socktop) at build time.
- Manifest adds CHOWN/DAC_OVERRIDE/FOWNER alongside SETUID/SETGID:
  with ALL dropped, uid 0 has no implicit file privilege and
  prepare_demo_home crash-looped on the demo-owned 700 home dir.
  Sessions still run with zero capabilities via setpriv.
- Agent liveness probe uses /proc instead of kill -0: without CAP_KILL
  even root gets EPERM signalling the socktop-user agent, so the old
  check false-alarmed in the pod logs.
- 0.3.11

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

.gitea/workflows/build-and-deploy.yaml

@@ -84,6 +84,25 @@ jobs:
           username: ${{ secrets.REGISTRY_USERNAME }}
           password: ${{ secrets.REGISTRY_PASSWORD }}
 
+      # Build into the runner's docker first (runner is arm64, so the image
+      # runs natively), gate on the CLI-compatibility check, and only then
+      # push. The layer cache makes the second build a no-op.
+      - name: Build Docker image (local, for verification)
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          file: ./Dockerfile
+          platforms: linux/arm64
+          push: false
+          load: true
+          tags: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:candidate
+          cache-from: type=registry,ref=${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:buildcache
+
+      - name: Verify installed socktop understands the shells' flags
+        run: |
+          chmod +x scripts/verify-image-socktop-flags.sh
+          scripts/verify-image-socktop-flags.sh ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:candidate
+
       - name: Build and push Docker image
         uses: docker/build-push-action@v5
         with:

Cargo.lock

@@ -2609,7 +2609,7 @@ dependencies = [
 
 [[package]]
 name = "webterm"
-version = "0.3.10"
+version = "0.3.11"
 dependencies = [
  "actix",
  "actix-files",

Cargo.toml

@@ -6,7 +6,7 @@ documentation = "https://docs.rs/webterm"
 readme = "README.md"
 categories = ["web-programming", "web-programming::websocket", "web-programming::http-server", "command-line-utilities"]
 keywords = ["terminal", "xterm", "websocket", "terminus", "console"]
-version = "0.3.10"
+version = "0.3.11"
 authors = ["fabian.freyer@physik.tu-berlin.de","jasonpwitty+socktop@proton.me"]
 edition = "2021"
 license = "BSD-3-Clause"

docker/entrypoint.sh

@@ -96,8 +96,10 @@ start_socktop_agent() {
     # Give it a moment to start
     sleep 1
 
-    # Check if it's running
-    if kill -0 $AGENT_PID 2>/dev/null; then
+    # Check if it's running. /proc, not kill -0: the agent runs as another UID
+    # and the pod's capability set strips CAP_KILL, so even root gets EPERM
+    # from a probe signal and the check would false-alarm.
+    if [ -d "/proc/$AGENT_PID" ]; then
         echo "  ✓ socktop-agent is running on port 3001"
     else
         echo "  ⚠ socktop-agent may have failed to start (check /tmp/socktop-agent.log)"

kubernetes/03-deployment.yaml

@@ -85,7 +85,7 @@ spec:
 
       containers:
         - name: webterm
-          image: gt.wittyoneoff.com/jason/socktop-webterm:0.3.10
+          image: gt.wittyoneoff.com/jason/socktop-webterm:0.3.11
           imagePullPolicy: Always
 
           command: ["/docker-entrypoint.sh"]
@@ -180,6 +180,15 @@ spec:
               add:
                 - SETUID
                 - SETGID
+                # prepare_demo_home (entrypoint.sh) writes into and re-owns
+                # /home/demo, which the image ships as demo-owned 700. With
+                # ALL dropped, uid 0 has no implicit file privilege, so the
+                # three file caps must come back or the entrypoint crashes on
+                # mkdir/chown/chmod. Sessions still get zero caps — session-
+                # shell.sh drops them all via setpriv --inh-caps -all.
+                - CHOWN
+                - DAC_OVERRIDE
+                - FOWNER
             readOnlyRootFilesystem: false
             runAsUser: 0
             runAsGroup: 0

scripts/verify-image-socktop-flags.sh

@@ -0,0 +1,54 @@
+#!/usr/bin/env bash
+# Verify that the socktop binary baked into a built image understands every
+# --flag the restricted/session shells pass to it.
+#
+# Why this exists: image 0.3.9 shipped with a cached apt layer holding socktop
+# 1.60.1, while restricted-shell.sh had started passing --no-kill (added in
+# 1.60.2). 1.60.1 parsed the unknown flag as the positional websocket URL,
+# prompted to overwrite the 'local' profile with url "--no-kill", and broke the
+# demo. This check fails the build whenever the shells and the installed
+# binary disagree about the CLI.
+#
+# Usage: verify-image-socktop-flags.sh IMAGE
+set -euo pipefail
+
+IMAGE="${1:?usage: verify-image-socktop-flags.sh IMAGE}"
+cd "$(dirname "$0")/.."
+
+# Every --flag appearing on a socktop invocation line in the shells.
+FLAGS=$(grep -hE '/usr/bin/socktop' docker/restricted-shell.sh docker/session-shell.sh 2>/dev/null |
+    grep -oE -- '--[a-z][a-z-]*' | sort -u)
+if [ -z "$FLAGS" ]; then
+    echo "ERROR: found no socktop flags to verify — did the shells move?" >&2
+    exit 1
+fi
+
+# The image is arm64-only; pin the platform so the check behaves the same on
+# the arm64 CI runner and on an amd64 box with qemu binfmt.
+run_socktop() {
+    docker run --rm --platform linux/arm64 --entrypoint /usr/bin/socktop "$IMAGE" "$@" 2>&1
+}
+
+if ! VERSION=$(run_socktop --version); then
+    echo "ERROR: could not run socktop from ${IMAGE}:" >&2
+    echo "$VERSION" >&2
+    exit 1
+fi
+HELP=$(run_socktop --help || true)
+echo "image socktop: ${VERSION}"
+
+rc=0
+for flag in $FLAGS; do
+    if printf '%s' "$HELP" | grep -q -- "$flag"; then
+        echo "  ok: $flag"
+    else
+        echo "  MISSING: installed socktop does not document $flag" >&2
+        rc=1
+    fi
+done
+
+if [ "$rc" -ne 0 ]; then
+    echo "FAIL: the image's socktop predates flags the shells pass." >&2
+    echo "Bump SOCKTOP_VERSION in the Dockerfile to a release that has them." >&2
+fi
+exit $rc