chore(quic-interop): drop the smoke target — runner is the canonical path

make smoke was the bisector for "is the bug in :quic or in the runner
environment?" while we were debugging the initial close-path / PTO /
padding issues. Now that the runner reliably runs the matrix end-to-end
(handshake + chacha20 green vs aioquic), smoke's only job — running
picoquic outside the runner — is unneeded, and the picoquic image
keeps changing its entrypoint / required args in ways that make smoke
finicky to maintain.

Removes:
  - make smoke / smoke-down targets
  - SMOKE_NET / SMOKE_PICOQUIC / SMOKE_CLIENT vars
  - SMOKE_MODE handling in run_endpoint.sh (just always tolerates
    /setup.sh failure now — same effect, less ceremony)
  - Plan doc note about make smoke updated to reflect removal.

If we ever need a non-runner bisector again, it's two `docker run`
commands; not worth permanent maintenance.

https://claude.ai/code/session_01HcvfQq1ttPV9PkRoJb4nyT

quic/interop/Makefile

@@ -1,21 +1,16 @@
 # Local iteration loop for the quic-interop-runner endpoint image.
 #
 #   make build              # compile JVM dist + build Docker image
-#   make smoke              # quick handshake test against a local picoquic
-#   make smoke-down         # tear down the smoke environment
 #   make image-name         # print the image tag (for use in implementations_quic.json)
+#   make clean              # nuke the dist + image
 #
-# These targets are deliberately thin wrappers — the real harness is the
-# `quic-interop-runner` repo (clone separately and register `amethyst` with
-# image=amethyst-quic-interop:latest, role=client).
-IMAGE          ?= amethyst-quic-interop:latest
-REPO_ROOT      := $(shell git rev-parse --show-toplevel)
-DIST_DIR       := build/install/quic-interop
-SMOKE_NET      := amethyst-quic-interop-smoke-net
-SMOKE_PICOQUIC := amethyst-quic-interop-smoke-picoquic
-SMOKE_CLIENT   := amethyst-quic-interop-smoke-client
+# Real testing happens via the runner: clone quic-interop-runner separately
+# and drive it with `quic/interop/run-matrix.sh -s <peer> -t <tests>`.
+IMAGE     ?= amethyst-quic-interop:latest
+REPO_ROOT := $(shell git rev-parse --show-toplevel)
+DIST_DIR  := build/install/quic-interop
 
-.PHONY: build dist docker smoke smoke-down image-name clean
+.PHONY: build dist docker image-name clean
 
 build: docker
 
@@ -25,36 +20,6 @@ dist:
 docker: dist
 	cd $(REPO_ROOT) && docker build -t $(IMAGE) -f quic/interop/Dockerfile .
 
-# Stand up picoquic + our endpoint on a private Docker bridge network. This
-# avoids `--network host`, which on Docker Desktop for Mac is misleadingly
-# *not* the Mac host's network — it's the LinuxKit VM's, and port forwarding
-# trickery makes 127.0.0.1 unreliable.
-#
-# We pass the resolved IP directly rather than relying on Docker's embedded
-# DNS — the base image (martenseemann/quic-network-simulator-endpoint) ships
-# a /etc/resolv.conf primed for the runner's sim that breaks bridge-network
-# name resolution from the JVM.
-smoke: docker smoke-down
-	docker network create $(SMOKE_NET) >/dev/null 2>&1 || true
-	docker run -d --name $(SMOKE_PICOQUIC) \
-	    --network $(SMOKE_NET) \
-	    --entrypoint /picoquic/picoquicdemo \
-	    privateoctopus/picoquic:latest -p 4433 >/dev/null
-	@PICOQUIC_IP=$$(docker inspect -f '{{range .NetworkSettings.Networks}}{{.IPAddress}}{{end}}' $(SMOKE_PICOQUIC)); \
-	    echo "picoquic at $$PICOQUIC_IP:4433 (in $(SMOKE_NET))"; \
-	    docker run --rm --name $(SMOKE_CLIENT) \
-	        --network $(SMOKE_NET) \
-	        -e SMOKE_MODE=1 \
-	        -e ROLE=client \
-	        -e TESTCASE=handshake \
-	        -e REQUESTS="https://$$PICOQUIC_IP:4433/" \
-	        $(IMAGE)
-	@$(MAKE) smoke-down
-
-smoke-down:
-	@docker rm -f $(SMOKE_PICOQUIC) $(SMOKE_CLIENT) >/dev/null 2>&1 || true
-	@docker network rm $(SMOKE_NET) >/dev/null 2>&1 || true
-
 image-name:
 	@echo $(IMAGE)
 

quic/interop/plans/2026-05-06-interop-runner.md

@@ -22,7 +22,10 @@ reorder / migration scenarios that are awkward to reproduce in unit tests.
   OpenJDK 21 runtime, copies the `installDist` output.
 - `run_endpoint.sh` sources the base image's `/setup.sh` then execs our JVM
   binary.
-- `Makefile` wrappers: `make build`, `make smoke`, `make clean`.
+- `Makefile` wrappers: `make build`, `make clean`. (A `make smoke`
+  target previously stood up picoquic + our endpoint on a private Docker
+  bridge to bisect runner failures from impl failures; dropped once
+  the runner reliably exercised both paths.)
 
 ## Local iteration loop
 

quic/interop/run_endpoint.sh

@@ -2,19 +2,14 @@
 # quic-interop-runner endpoint entry point.
 #
 # The base image (martenseemann/quic-network-simulator-endpoint) provides
-# /setup.sh, which configures routing for the runner's ns-3 sim. We source
-# it inside the runner — but for `make smoke` runs against a plain Docker
-# bridge network, sourcing /setup.sh succeeds AND breaks DNS resolution
-# (it points the resolver at the sim's nameserver which doesn't exist
-# in smoke mode). Set SMOKE_MODE=1 to skip /setup.sh entirely.
+# /setup.sh, which configures routing for the runner's ns-3 sim. Source it
+# inside the runner; tolerate failure (e.g. missing NET_ADMIN cap) so the
+# JVM client still launches if a caller invokes the image outside the
+# runner — they get default container networking instead.
 set -uo pipefail
 
-if [ "${SMOKE_MODE:-0}" = "1" ]; then
-    echo "(smoke mode — skipping /setup.sh, using container default networking)" >&2
-else
-    # shellcheck disable=SC1091
-    source /setup.sh 2>/dev/null || echo "(setup.sh skipped — running outside the runner sim)" >&2
-fi
+# shellcheck disable=SC1091
+source /setup.sh 2>/dev/null || echo "(setup.sh skipped — not under the runner sim)" >&2
 
 set -e