fix: deploy step stops containers before recreating to avoid ContainerConfig bug

docker-compose 1.29.x fails with KeyError: 'ContainerConfig' when trying
to recreate containers built from newer image manifests. Explicitly stopping
and removing before 'up -d' avoids the inspection of old container state.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

.gitea/workflows/docker-build.yml

@@ -128,13 +128,22 @@ jobs:
           script: |
             cd ${{ secrets.DEPLOY_PATH }}
             echo "${{ secrets.REGISTRY_TOKEN }}" | docker login git.necloud.us -u "${{ secrets.REGISTRY_USER }}" --password-stdin
+
+            # Pull new images first
             if docker compose version > /dev/null 2>&1; then
-              docker compose pull backend frontend
-              docker compose up -d --no-build backend frontend
+              COMPOSE="docker compose"
             else
-              docker-compose pull backend frontend
-              docker-compose up -d --no-build backend frontend
+              COMPOSE="docker-compose"
             fi
+
+            $COMPOSE pull backend frontend
+
+            # Stop and remove old containers so docker-compose up creates fresh ones
+            # (works around ContainerConfig bug in docker-compose 1.29.x)
+            $COMPOSE stop backend frontend || true
+            $COMPOSE rm -f backend frontend || true
+            $COMPOSE up -d backend frontend
+
             docker logout git.necloud.us
             docker image prune -f