revert(core): revert malformed streamed response retry logic - #6783
Merged
Conversation
This reverts commit 43d34f7.
wenshao
changed the title
qwen-code-ci-bot
previously requested changes
Jul 12, 2026
qwen-code-ci-bot
left a comment
qwen-code-ci-bot
left a comment
Collaborator
There was a problem hiding this comment.
This is a revert of #6754 (retry malformed streamed responses), which was merged just yesterday. The PR body is a one-liner — "Reverts #6754" — with no template headings and, more importantly, no explanation of why the fix is being backed out.
A few things I noticed while looking at the context:
- The original PR addressed #6666, a real production bug where thinking tags leaked into visible content. That issue is currently closed (closed by @wenshao on 2026-07-12).
- @yiliang114 (the original PR author) also reverted their own related PR #6751 the same day — "Preserve tagged thinking parsing for Qwen3 default provider." That's a clear pattern of backing out the client-side thinking-tag detection approach.
- The revert itself is clean — the split retry budget (
transientMaxRetries: 4/protocolTagLeakMaxRetries: 2) is correctly restored to its pre-#6754 state, and all converter-side detection logic, the separateInvalidStreamErrorfile, and the nameless-tool-call handling are fully removed.
The question is: what changed since yesterday that makes this revert the right call? Was the bug fixed server-side? Did #6754 cause regressions? The PR body doesn't say, and that context matters — both for the review and for anyone who hits #6666 again in the future.
Could you add a short explanation to the PR body? Something like:
- Why the revert is needed (server-side fix? regression? approach change?)
- Whether #6666 is still a risk or if it's resolved through another mechanism
中文说明
这是对 #6754（重试 malformed streamed responses）的 revert，而 #6754 昨天才合并。PR body 只有一行"Reverts #6754"，没有模板段落，更重要的是没有解释为什么要撤回这个 fix。
几个背景：
- 原始 PR 修复的是 #6666，一个真实的生产 bug（thinking tags 泄漏到 visible content）。该 issue 目前状态为 closed（由 @wenshao 在 2026-07-12 关闭）。
- @yiliang114（原始 PR 作者）同一天也 revert 了自己相关的 PR #6751（"Preserve tagged thinking parsing for Qwen3 default provider"）。明显是在系统性地撤回客户端 thinking-tag 检测方案。
- Revert 本身是干净的——分离重试预算（
transientMaxRetries: 4/protocolTagLeakMaxRetries: 2）正确恢复到了 #6754 之前的状态，converter 侧的检测逻辑、独立的InvalidStreamError文件、nameless-tool-call 处理都被完整移除。
问题是：从昨天到现在发生了什么变化，使得 revert 成为正确选择？ bug 是在服务端修复了？还是 #6754 引入了回归？PR body 没有说明，这个上下文对 review 和未来遇到 #6666 的人都很重要。
能否在 PR body 加一段简短说明：
- 为什么需要 revert（服务端修复？回归？方案变更？）
- #6666 是否还有风险，还是已经通过其他机制解决了？
— Qwen Code · qwen3.7-max
wenshao
dismissed
qwen-code-ci-bot’s stale review
July 12, 2026 22:52
PR body has been updated with full regression analysis: E2E went from 0 to ~14 failed files / 31 failed tests after #6754 merged. Detection heuristics (hasVisibleThinkingTagLeakSignature, shouldDropMalformedAttempt) misclassify legitimate responses as malformed, silently dropping tool calls. Issue #6666 should be re-addressed with less aggressive heuristics in a follow-up PR.
Contributor
Code Coverage Summary
CLI Package - Full Text ReportCore Package - Full Text ReportFor detailed HTML reports, please see the 'coverage-reports-22.x-ubuntu-latest' artifact from the main CI run. |
qwen-code-ci-bot
approved these changes
Jul 12, 2026
qwen-code-ci-bot
left a comment
qwen-code-ci-bot
left a comment
Collaborator
There was a problem hiding this comment.
No issues found. LGTM! ✅
— qwen3.7-max via Qwen Code /review
pomelo-nwu
approved these changes
Jul 12, 2026
This was referenced Jul 13, 2026
pull Bot
pushed a commit
to mcx/qwen-code
that referenced
this pull request
Jul 14, 2026
…ction (QwenLM#6794) * Revert "Revert "fix(core): retry malformed streamed responses (QwenLM#6754)" (QwenLM#6783)" This reverts commit ff7d48a. * fix(core): narrow nameless tool-call detection to stop false-positive drops hasNamelessToolCall() flagged any buffer slot lacking a name, including phantom slots (no id, no argument content) created by trailing structural deltas. That tripped shouldDropMalformedAttempt on legitimate tool-call responses, clearing parts and suppressing finish_reason to undefined, which surfaced as 'Model stream ended without a finish reason' and broke E2E suites depending on tool-call emission. Only treat a nameless tool call as malformed when the provider actually started a real call (assigned id or streamed argument content). Genuine QwenLM#6666 nameless attempts (id or args but no name) still drop and retry; phantom slots are skipped by getCompletedToolCalls as before. * fix(core): preserve thinking tag stream boundaries * fix(core): address malformed stream review feedback * fix(core): tighten malformed stream retry signals * fix(core): address remaining stream review feedback * test(core): cover stream review edge cases * fix(core): address stream review regressions * refactor(core): remove duplicate thinking tag state * fix(core): close reverse-order stream validation gaps * fix(core): preserve cross-channel tag evidence * test(core): cover split cross-channel tags * fix(core): close remaining stream routing gaps * fix(core): handle late tool call names * fix(core): preserve active tool call routes * test(core): cover malformed stream edge cases * fix(core): preserve streaming tool call routes * refactor(core): simplify malformed stream validation * docs(core): clarify invalid stream retry budget * fix(core): preserve valid streamed metadata boundaries * fix(core): deduplicate streamed tool metadata * fix(core): preserve streamed protocol line boundaries
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this PR does
Reverts PR #6754 (
fix(core): retry malformed streamed responses), which introduced aggressive thinking-tag leak detection and nameless tool-call dropping logic in the OpenAI→Gemini stream converter. The reverted changes includehasVisibleThinkingTagLeakSignature(),pendingUntrustedResponsePartsbuffering,shouldDropMalformedAttemptsuppression offinish_reason, and thehasNamelessToolCall()method onStreamingToolCallParser.Why it's needed
PR #6754 caused a widespread E2E regression on
main. After it was merged (43d34f739d), the E2E suite went from 0 failures to ~14 failed test files / 31 failed tests across both Linux and macOS. The stream converter's new detection heuristics misclassify legitimate responses as malformed, causing tool calls to be silently dropped (parts.length = 0,finish_reasonsuppressed toundefined). All SDK and CLI E2E tests that depend on tool-call emission are affected —permission-control,tool-control,sdk-mcp-server,subagents,multi-turn,edit,file-system,write_file,utf-bom-encoding,cron-tools,notebook-edit,tool-search, etc. The malformed response retry feature should be re-implemented with less aggressive detection heuristics.Reviewer Test Plan
How to verify
cd packages/core && npx vitest run src/core/openaiContentGenerator/converter.test.ts src/core/geminiChat.test.tsEvidence (Before & After)
N/A — non-UI revert.
Tested on
Environment (optional)
CI E2E runs.
Risk & Scope
Linked Issues
Reverts #6754
中文说明
本 PR 做了什么
回滚 PR #6754（
fix(core): retry malformed streamed responses）。该 PR 在 OpenAI→Gemini 流式转换器中引入了激进的 thinking 标签泄漏检测和无名工具调用丢弃逻辑，包括hasVisibleThinkingTagLeakSignature()、pendingUntrustedResponseParts缓冲、shouldDropMalformedAttempt抑制finish_reason，以及StreamingToolCallParser上的hasNamelessToolCall()方法。为什么需要
PR #6754 在
main分支上引发了大面积 E2E 回归。合入后（43d34f739d），E2E 测试从 0 失败变为 ~14 个文件失败 / 31 个用例失败，Linux 和 macOS 均受影响。流式转换器的新检测启发式算法将合法响应误判为畸形，导致工具调用被静默丢弃（parts.length = 0，finish_reason被抑制为undefined）。所有依赖工具调用发射的 SDK 和 CLI E2E 测试均受影响。畸形响应重试功能应以更保守的检测策略重新实现。风险与范围