cli: run duplicates the message when passed after -- #57678

Activity

1. opencode-agent commented on Oct 21, 2026

   

   opencode-agentBot

   on Oct 21, 2026 – with OpenCode Agent

   Contributor

   More actions

   Thanks for the clear report. I could reproduce this on 2.1.2 and it still happens on the latest dev (4d59657).

   I used a local mock OpenAI-compatible provider that logs the user message it receives:

   • opencode run --model mock/mock DUPREPRO sends "DUPREPRO", which is correct.
   • opencode run --model mock/mock -- DUPREPRO sends "DUPREPRO DUPREPRO", both as the turn message and in the title-generation prompt. The command still exits 0.

   The script below reproduces it. My guess, not yet confirmed, is that the run handler adds the arguments after -- to the message again, even though the argument parser already includes them.

   The second symptom, a single argument with spaces getting wrapped in literal quotes, also reproduces. It's already tracked in #56120.

   Reproduction script (repro.sh)

   #!/usr/bin/env bash
   # Reproduces #57678: `opencode run -- MSG` sends/stores "MSG MSG".
   # Needs bun and @opencode/cli on PATH (bun add -g @opencode/[email protected]).
   # A local OpenAI-compatible mock prints the user message it receives.
   set -euo pipefail
   W=$(mktemp -d /tmp/oc-dup.XXXXXX)
   export HOME="$W/home"; mkdir -p "$HOME/.config/opencode" "$W/proj"
   git init -q "$W/proj"
   
   cat > "$W/mock.ts" <<'EOF'
   Bun.serve({ port: 4555, async fetch(req) {
     if (new URL(req.url).pathname.endsWith("/models")) return Response.json({ data: [{ id: "mock" }] })
     const body = await req.json()
     const last = body.messages.filter((m: any) => m.role === "user").at(-1)
     console.log("USER:", JSON.stringify(last?.content))
     const chunk = (d: any) => `data: ${JSON.stringify({ id: "x", object: "chat.completion.chunk", created: 0, model: "mock", choices: [d] })}\n\n`
     return new Response(chunk({ index: 0, delta: { role: "assistant", content: "ok" } }) + chunk({ index: 0, delta: {}, finish_reason: "stop" }) + "data: [DONE]\n\n",
       { headers: { "content-type": "text/event-stream" } })
   }})
   EOF
   bun "$W/mock.ts" > "$W/mock.log" 2>&1 & MOCK=$!
   trap 'kill $MOCK 2>/dev/null' EXIT
   sleep 1
   
   cat > "$HOME/.config/opencode/opencode.jsonc" <<'EOF'
   { "providers": { "mock": { "package": "@opencode/ai/providers/openai-compatible",
     "settings": { "baseURL": "http://127.0.0.1:4555/v1", "apiKey": "x" },
     "models": { "mock": { "name": "mock" } } } } }
   EOF
   
   cd "$W/proj"
   for args in "DUPREPRO" "-- DUPREPRO"; do
     : > "$W/mock.log"
     opencode run --standalone --format json --model mock/mock $args </dev/null > /dev/null
     echo "== opencode run ... $args"
     grep -av "first message of the conversation" "$W/mock.log" | grep -a USER: | sed 's/^/   /' | head -1
   done
   # Expected: both print USER: "DUPREPRO". Actual: the `--` run prints USER: "DUPREPRO DUPREPRO".

   Run it: bash repro.sh
2. opencode-agent

   added

   reproduced

   on Oct 21, 2026
3. Ethereal49 commented on Oct 21, 2026

   

   Ethereal49

   on Oct 21, 2026

   More actions

   I’ll trace the run argument handling and fix the duplicate message for the -- case, keeping the separate literal-quote symptom out of scope. I’ll add regression coverage for both invocations with a local mock provider and report the exact verification boundary. Codex is assisting with the investigation and implementation.
4. rekram1-node

   closed this as completedin #61721on Nov 4, 2026