# Reasoning-block trailing newline parser bug causing agent loops > The loop was reported as llama.cpp issue 24181, titled "Eval bug: Step 3.7 Flash gets stuck in reasoning trying to make tool calls (autoparser)", renamed on 2026-06-05 from "Bug: Step 3.7 Flash enters infinite reasoning loops". Parent: [Mac local LLMs: Chat templates, reasoning and tool calling](https://llms-explorer.com/tree/mac-local-llms-chat-templates-reasoning-and-tool-calling/) · 1 facets · 12 facts · page: https://llms-explorer.com/tree/reasoning-block-trailing-newline-parser-bug-caus/ ## Facts - The loop was reported as llama.cpp issue 24181, titled "Eval bug: Step 3.7 Flash gets stuck in reasoning trying to make tool calls (autoparser)", renamed on 2026-06-05 from "Bug: Step 3.7 Flash enters infinite reasoning loops". — [source](https://github.com/ggml-org/llama.cpp/issues/24181) - A Hacker News commenter who debugged it says the parser captured an extra `\n` as part of the reasoning block, it appeared only in longer multi-turn agentic sessions, and the extra linefeed steered the model into reasoning self-corrections that worsened with session length. — [source](https://news.ycombinator.com/item?id=49402232) - A second commenter's reading of the issue thread: a `\n` slipped through from the last line of the reasoning trace, producing a blank line before `` that occasionally triggered an "Actually ..." digression which then looped. — [source](https://news.ycombinator.com/item?id=49402232) - The first commenter explains the loop as incremental in-context imitation: a model trained to end reasoning with one linefeed then `` sees two consecutive linefeeds in its replayed history, adds one itself, and each digression raises the chance of the next. — [source](https://news.ycombinator.com/item?id=49402232) - The same commenter says the autoparser "was (and still is)" parsing a trailing linefeed as part of a reasoning block and that a complete fix needs both a parser-definition fix and whitespace trimming in the encoding step, since the latter also covers clients that add the whitespace deliberately. — [source](https://news.ycombinator.com/item?id=49402232) - He states the maintainer fixed this specific issue by trimming the extra linefeed before passing the text to the template. — [source](https://news.ycombinator.com/item?id=49402232) - He describes the autoparser's goal as inferring delimiter tokens from the chat template so new model parsers are easier to write, and notes that llama.cpp's API returns pre-parsed data so clients need not parse thinking, text or tool blocks. — [source](https://news.ycombinator.com/item?id=49402232) - In issue 24181 the autoparser found raw reasoning delimiters with surrounding whitespace (start `\n`, end `\n\n`) while `common_chat_params` exposed only the trimmed tags. — [source](https://github.com/ggml-org/llama.cpp/issues/24181) - A patch posted in issue 24181 opts Step 3.7 Flash out of the autoparser with a dedicated parser: the generation prompt is `<|im_start|>assistant\n\n` plus reasoning, `\n\n` and content; the open delimiter is `\n` or `` with an optional newline; the close is an optional `\n` before `` and an optional `\n` after; and reasoning text is read up to `\n` or ``. — [source](https://github.com/ggml-org/llama.cpp/issues/24181) - The same patch uses a lazy grammar triggered by the word `` when tools are present and tool choice is auto. — [source](https://github.com/ggml-org/llama.cpp/issues/24181) - Inferred: the delimiter's newline must belong to the delimiter (consumed by the parser) and not to the reasoning text, otherwise every round trip through a non-trimming template adds a line feed. — source: `asserted` - Inferred: a server-side regression test should replay a reasoning turn through parse then template and assert the rendered bytes equal the generated bytes. — source: `asserted`