Skip to content

merge queue: checking release/v3.8.49 (7123236), #7042, #7093, #7098 and #7102 together - #7391

Closed
mergify[bot] wants to merge 14 commits into
release/v3.8.49from
mergify/merge-queue/129251e97f
Closed

mergify[bot] wants to merge 14 commits into
release/v3.8.49from
mergify/merge-queue/129251e97f

Conversation

@mergify

@mergify mergify Bot commented Jul 16, 2026 •

Copy link
Copy Markdown

✨ Pull request #7042 ahead in the queue was removed (reason: checks failed). The pull request #7102 has been requeued. ✨

Branch release/v3.8.49 (7123236), #7042, #7093, #7098 and #7102 are queued together for merge.

This pull request has been created by Mergify to speculatively check the mergeability of #7102.
You don't need to do anything. Mergify will close this pull request automatically when it is complete.

Required conditions of queue rule release for merge:

  • #check-pending=0
  • #check-success>=1
  • check-success=Merge integrity (changelog + generated skills)
  • any of:
    • #check-failure=0
    • all of:
      • #check-failure=1
      • check-failure=dast-smoke

Required conditions to stay in the queue:

---
checking_base_sha: e132a1582d34f9f9407c09cc52461003da1f1cce
previous_check_retries: []
previous_failed_batches: []
pull_requests:
  - number: 7102
    scopes: []
scopes: []
...

alltomatos and others added 14 commits July 13, 2026 09:47
…ort from 9router#1321)

The reasoning_content injector already handles DeepSeek/Kimi/K2/MiniMax thinking-mode upstreams, echoing a placeholder reasoning_content on assistant turns that lack one. Its THINKING_MODEL_PATTERNS list omitted the xiaomi-tokenplan mimo family, so requests through xiaomi-tokenplan/mimo-v2.5-pro still hit upstream's 400 'reasoning_content in the thinking mode must be passed back to the API', making the model unusable in multi-turn conversations (e.g. Codex CLI). Add a /\bmimo\b/i pattern so mimo models get the same treatment.

Reported-by: z.wl (@xxue-z) (decolua/9router#1321)
… (port from 9router#2132)

Root cause: SmartCrusher's system-message guard only excluded role === "system", but Codex CLI (open-sse/executors/codex.ts) sends its instructions/tool-schema turn with role "developer" (the Responses-API equivalent of system used by newer models). Every other system-exclusion guard in this codebase also covers developer (roleNormalizer.ts, contextManager.ts, claudeUpstreamMessages.ts, etc.) except this one, so Headroom happily tabular-compacted JSON arrays embedded in the developer turn (e.g. an update_plan tool schema example), corrupting the instructions the model needs to call the plan tool and breaking Codex CLI plan mode.

Fix: extend the guard in crushMessages()/collectCompactableArrays() (smartcrusher.ts) to skip role === "developer" alongside role === "system".

Reported-by: SingCJ (decolua/9router#2132)
@mergify mergify Bot closed this Jul 16, 2026
@mergify
mergify Bot deleted the mergify/merge-queue/129251e97f branch July 16, 2026 10:28
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants