fix(cli): recover leaked mouse tracking terminal state - #17701
Merged
Conversation
Detect leaked SGR mouse-report fragments in CLI input, strip them, and reset terminal modes in-place so scroll and typing recover without reopening the tab. Add regression tests for escaped, visible, and bare leak forms.
Reset sticky mouse/focus/paste terminal modes before the TUI starts and during graceful shutdown paths so stale tab state from prior crashes cannot poison the next session.
Contributor
There was a problem hiding this comment.
Pull request overview
Extends the CLI’s defensive input sanitization to handle leaked SGR mouse-report fragments and adds a best-effort terminal mode recovery when those leaks are detected, backed by regression tests.
Changes:
- Add SGR mouse report stripping (ESC / visible
^[[/ bare<...M/mfragments) to the leaked-terminal response sanitizer. - Introduce a “with meta” sanitizer variant returning
(cleaned_text, had_mouse_reports)and trigger an in-place terminal mode reset when mouse leaks are detected. - Add regression tests covering escaped, visible, bare, and concatenated mouse-report leak forms while preserving normal angle-bracket text.
Reviewed changes
Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
cli.py |
Adds regexes for SGR mouse leaks, returns meta from sanitizer, and performs terminal input-mode recovery when leaks are detected. |
tests/cli/test_cli_terminal_response_sanitizer.py |
Adds test coverage for stripping multiple forms of leaked SGR mouse-report fragments without over-matching normal <...> text. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
alt-glitch
added
type/bug
Handle unbounded SGR mouse report coordinates and avoid regex work on ordinary prompt-buffer edits by short-circuiting before sanitizer passes.
Contributor
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Contributor
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+1583
to
+1598
| if not any(marker in text for marker in _TERMINAL_RESPONSE_SENTINELS): | ||
| return text, False | ||
|
|
||
| had_mouse_reports = False | ||
|
|
||
| if "\x1b[" in text: | ||
| text = _DSR_CPR_ESC_RE.sub("", text) | ||
| text, count = _SGR_MOUSE_ESC_RE.subn("", text) | ||
| had_mouse_reports = had_mouse_reports or count > 0 | ||
|
|
||
| if "^[" in text: | ||
| text = _DSR_CPR_VISIBLE_RE.sub("", text) | ||
| text, count = _SGR_MOUSE_VISIBLE_RE.subn("", text) | ||
| had_mouse_reports = had_mouse_reports or count > 0 | ||
|
|
||
| if "<" in text: |
There was a problem hiding this comment.
In _strip_leaked_terminal_responses_with_meta, the sentinel fast-path is likely to be bypassed for a lot of normal developer input because _TERMINAL_RESPONSE_SENTINELS includes "<" (common in code/HTML/generics). Also, because "<" is last in the tuple, any(marker in text ...) will scan the entire buffer for "\x1b[" and "^[" before it ever checks for <, and the function then repeats additional in scans in the per-marker blocks. Since this runs from _on_text_changed (per-keystroke), consider restructuring the fast-path to avoid multiple full-string scans (e.g., compute has_esc/has_visible/has_bare once and early-return if all false, and/or make the bare check more specific like requiring ; + M/m before running the bare regex).
Suggested change
| if not any(marker in text for marker in _TERMINAL_RESPONSE_SENTINELS): | |
| return text, False | |
| had_mouse_reports = False | |
| if "\x1b[" in text: | |
| text = _DSR_CPR_ESC_RE.sub("", text) | |
| text, count = _SGR_MOUSE_ESC_RE.subn("", text) | |
| had_mouse_reports = had_mouse_reports or count > 0 | |
| if "^[" in text: | |
| text = _DSR_CPR_VISIBLE_RE.sub("", text) | |
| text, count = _SGR_MOUSE_VISIBLE_RE.subn("", text) | |
| had_mouse_reports = had_mouse_reports or count > 0 | |
| if "<" in text: | |
| has_esc = "\x1b[" in text | |
| has_visible = "^[" in text | |
| has_mouse_suffix = "M" in text or "m" in text | |
| has_bare_mouse = "<" in text and ";" in text and has_mouse_suffix | |
| if not (has_esc or has_visible or has_bare_mouse): | |
| return text, False | |
| had_mouse_reports = False | |
| if has_esc: | |
| text = _DSR_CPR_ESC_RE.sub("", text) | |
| text, count = _SGR_MOUSE_ESC_RE.subn("", text) | |
| had_mouse_reports = had_mouse_reports or count > 0 | |
| if has_visible: | |
| text = _DSR_CPR_VISIBLE_RE.sub("", text) | |
| text, count = _SGR_MOUSE_VISIBLE_RE.subn("", text) | |
| had_mouse_reports = had_mouse_reports or count > 0 | |
| if has_bare_mouse: |
02356abc
pushed a commit
to 02356abc/hermes-agent
that referenced
this pull request
May 14, 2026
…-self-heal fix(cli): recover leaked mouse tracking terminal state
jsboige
pushed a commit
to jsboige/hermes-agent
that referenced
this pull request
May 14, 2026
…-self-heal fix(cli): recover leaked mouse tracking terminal state
dannyJ848
pushed a commit
to dannyJ848/hermes-agent
that referenced
this pull request
May 17, 2026
…-self-heal fix(cli): recover leaked mouse tracking terminal state
This was referenced May 20, 2026
gweeteve
pushed a commit
to gweeteve/hermes-agent
that referenced
this pull request
Jun 2, 2026
…-self-heal fix(cli): recover leaked mouse tracking terminal state
Merged
4 tasks
OutThisLife
added a commit
that referenced
this pull request
Jun 4, 2026
With the tokenizer reassembling split CSI sequences across a flush (prior commit), no SGR mouse fragment can reach a text token anymore — terminals write a mouse report as one atomic sequence, and any read/flush split now re-joins in the tokenizer buffer instead of leaking. That makes the whole downstream recovery layer dead code: - SGR_MOUSE_FRAGMENT_RE, MOUSE_BURST_NOISE_RE, MOUSE_BURST_RESIDUE_RE - parseTextWithSgrMouseFragments / parseSgrMouseFragment / normalizeSgrMouseFragment - the whole-text mouse-burst noise fast path in parseMultipleKeypresses Remove all of it (~185 lines) and the tests that only exercised it. The narrow legacy X10 wheel-tail resynth stays (distinct mechanism, kept with its own test). This retires the #17701 → #18113 → #26781 → #28463 → #35512 regex hardening chain in favor of the one correct parser fix.
Yuki-14544869
pushed a commit
to Yuki-14544869/hermes-agent
that referenced
this pull request
Jun 4, 2026
With the tokenizer reassembling split CSI sequences across a flush (prior commit), no SGR mouse fragment can reach a text token anymore — terminals write a mouse report as one atomic sequence, and any read/flush split now re-joins in the tokenizer buffer instead of leaking. That makes the whole downstream recovery layer dead code: - SGR_MOUSE_FRAGMENT_RE, MOUSE_BURST_NOISE_RE, MOUSE_BURST_RESIDUE_RE - parseTextWithSgrMouseFragments / parseSgrMouseFragment / normalizeSgrMouseFragment - the whole-text mouse-burst noise fast path in parseMultipleKeypresses Remove all of it (~185 lines) and the tests that only exercised it. The narrow legacy X10 wheel-tail resynth stays (distinct mechanism, kept with its own test). This retires the NousResearch#17701 → NousResearch#18113 → NousResearch#26781 → NousResearch#28463 → NousResearch#35512 regex hardening chain in favor of the one correct parser fix.
davidgut1982
pushed a commit
to davidgut1982/hermes-agent
that referenced
this pull request
Jun 5, 2026
With the tokenizer reassembling split CSI sequences across a flush (prior commit), no SGR mouse fragment can reach a text token anymore — terminals write a mouse report as one atomic sequence, and any read/flush split now re-joins in the tokenizer buffer instead of leaking. That makes the whole downstream recovery layer dead code: - SGR_MOUSE_FRAGMENT_RE, MOUSE_BURST_NOISE_RE, MOUSE_BURST_RESIDUE_RE - parseTextWithSgrMouseFragments / parseSgrMouseFragment / normalizeSgrMouseFragment - the whole-text mouse-burst noise fast path in parseMultipleKeypresses Remove all of it (~185 lines) and the tests that only exercised it. The narrow legacy X10 wheel-tail resynth stays (distinct mechanism, kept with its own test). This retires the NousResearch#17701 → NousResearch#18113 → NousResearch#26781 → NousResearch#28463 → NousResearch#35512 regex hardening chain in favor of the one correct parser fix.
changman
pushed a commit
to changman/hermes-agent
that referenced
this pull request
Jun 10, 2026
With the tokenizer reassembling split CSI sequences across a flush (prior commit), no SGR mouse fragment can reach a text token anymore — terminals write a mouse report as one atomic sequence, and any read/flush split now re-joins in the tokenizer buffer instead of leaking. That makes the whole downstream recovery layer dead code: - SGR_MOUSE_FRAGMENT_RE, MOUSE_BURST_NOISE_RE, MOUSE_BURST_RESIDUE_RE - parseTextWithSgrMouseFragments / parseSgrMouseFragment / normalizeSgrMouseFragment - the whole-text mouse-burst noise fast path in parseMultipleKeypresses Remove all of it (~185 lines) and the tests that only exercised it. The narrow legacy X10 wheel-tail resynth stays (distinct mechanism, kept with its own test). This retires the NousResearch#17701 → NousResearch#18113 → NousResearch#26781 → NousResearch#28463 → NousResearch#35512 regex hardening chain in favor of the one correct parser fix.
Seven74AI
pushed a commit
to Seven74AI/hermes-agent
that referenced
this pull request
Jun 13, 2026
…-self-heal fix(cli): recover leaked mouse tracking terminal state
alt-glitch
pushed a commit
that referenced
this pull request
Jun 14, 2026
With the tokenizer reassembling split CSI sequences across a flush (prior commit), no SGR mouse fragment can reach a text token anymore — terminals write a mouse report as one atomic sequence, and any read/flush split now re-joins in the tokenizer buffer instead of leaking. That makes the whole downstream recovery layer dead code: - SGR_MOUSE_FRAGMENT_RE, MOUSE_BURST_NOISE_RE, MOUSE_BURST_RESIDUE_RE - parseTextWithSgrMouseFragments / parseSgrMouseFragment / normalizeSgrMouseFragment - the whole-text mouse-burst noise fast path in parseMultipleKeypresses Remove all of it (~185 lines) and the tests that only exercised it. The narrow legacy X10 wheel-tail resynth stays (distinct mechanism, kept with its own test). This retires the #17701 → #18113 → #26781 → #28463 → #35512 regex hardening chain in favor of the one correct parser fix.
kossteg
pushed a commit
to kossteg/hermes-agent
that referenced
this pull request
Jun 16, 2026
With the tokenizer reassembling split CSI sequences across a flush (prior commit), no SGR mouse fragment can reach a text token anymore — terminals write a mouse report as one atomic sequence, and any read/flush split now re-joins in the tokenizer buffer instead of leaking. That makes the whole downstream recovery layer dead code: - SGR_MOUSE_FRAGMENT_RE, MOUSE_BURST_NOISE_RE, MOUSE_BURST_RESIDUE_RE - parseTextWithSgrMouseFragments / parseSgrMouseFragment / normalizeSgrMouseFragment - the whole-text mouse-burst noise fast path in parseMultipleKeypresses Remove all of it (~185 lines) and the tests that only exercised it. The narrow legacy X10 wheel-tail resynth stays (distinct mechanism, kept with its own test). This retires the NousResearch#17701 → NousResearch#18113 → NousResearch#26781 → NousResearch#28463 → NousResearch#35512 regex hardening chain in favor of the one correct parser fix.
T02200059
pushed a commit
to T02200059/hermes-agent
that referenced
this pull request
Jun 18, 2026
With the tokenizer reassembling split CSI sequences across a flush (prior commit), no SGR mouse fragment can reach a text token anymore — terminals write a mouse report as one atomic sequence, and any read/flush split now re-joins in the tokenizer buffer instead of leaking. That makes the whole downstream recovery layer dead code: - SGR_MOUSE_FRAGMENT_RE, MOUSE_BURST_NOISE_RE, MOUSE_BURST_RESIDUE_RE - parseTextWithSgrMouseFragments / parseSgrMouseFragment / normalizeSgrMouseFragment - the whole-text mouse-burst noise fast path in parseMultipleKeypresses Remove all of it (~185 lines) and the tests that only exercised it. The narrow legacy X10 wheel-tail resynth stays (distinct mechanism, kept with its own test). This retires the NousResearch#17701 → NousResearch#18113 → NousResearch#26781 → NousResearch#28463 → NousResearch#35512 regex hardening chain in favor of the one correct parser fix.
waefrebeorn
pushed a commit
to waefrebeorn/slermes
that referenced
this pull request
Jul 2, 2026
…-self-heal fix(cli): recover leaked mouse tracking terminal state
waefrebeorn
pushed a commit
to waefrebeorn/slermes
that referenced
this pull request
Jul 2, 2026
With the tokenizer reassembling split CSI sequences across a flush (prior commit), no SGR mouse fragment can reach a text token anymore — terminals write a mouse report as one atomic sequence, and any read/flush split now re-joins in the tokenizer buffer instead of leaking. That makes the whole downstream recovery layer dead code: - SGR_MOUSE_FRAGMENT_RE, MOUSE_BURST_NOISE_RE, MOUSE_BURST_RESIDUE_RE - parseTextWithSgrMouseFragments / parseSgrMouseFragment / normalizeSgrMouseFragment - the whole-text mouse-burst noise fast path in parseMultipleKeypresses Remove all of it (~185 lines) and the tests that only exercised it. The narrow legacy X10 wheel-tail resynth stays (distinct mechanism, kept with its own test). This retires the NousResearch#17701 → NousResearch#18113 → NousResearch#26781 → NousResearch#28463 → NousResearch#35512 regex hardening chain in favor of the one correct parser fix.
santhreal
pushed a commit
to santhreal/hermes-agent
that referenced
this pull request
Jul 13, 2026
With the tokenizer reassembling split CSI sequences across a flush (prior commit), no SGR mouse fragment can reach a text token anymore — terminals write a mouse report as one atomic sequence, and any read/flush split now re-joins in the tokenizer buffer instead of leaking. That makes the whole downstream recovery layer dead code: - SGR_MOUSE_FRAGMENT_RE, MOUSE_BURST_NOISE_RE, MOUSE_BURST_RESIDUE_RE - parseTextWithSgrMouseFragments / parseSgrMouseFragment / normalizeSgrMouseFragment - the whole-text mouse-burst noise fast path in parseMultipleKeypresses Remove all of it (~185 lines) and the tests that only exercised it. The narrow legacy X10 wheel-tail resynth stays (distinct mechanism, kept with its own test). This retires the NousResearch#17701 → NousResearch#18113 → NousResearch#26781 → NousResearch#28463 → NousResearch#35512 regex hardening chain in favor of the one correct parser fix.
donbowman
pushed a commit
to donbowman/hermes-agent
that referenced
this pull request
Jul 13, 2026
With the tokenizer reassembling split CSI sequences across a flush (prior commit), no SGR mouse fragment can reach a text token anymore — terminals write a mouse report as one atomic sequence, and any read/flush split now re-joins in the tokenizer buffer instead of leaking. That makes the whole downstream recovery layer dead code: - SGR_MOUSE_FRAGMENT_RE, MOUSE_BURST_NOISE_RE, MOUSE_BURST_RESIDUE_RE - parseTextWithSgrMouseFragments / parseSgrMouseFragment / normalizeSgrMouseFragment - the whole-text mouse-burst noise fast path in parseMultipleKeypresses Remove all of it (~185 lines) and the tests that only exercised it. The narrow legacy X10 wheel-tail resynth stays (distinct mechanism, kept with its own test). This retires the NousResearch#17701 → NousResearch#18113 → NousResearch#26781 → NousResearch#28463 → NousResearch#35512 regex hardening chain in favor of the one correct parser fix.
Gravezzz
pushed a commit
to Gravezzz/hermes-agent
that referenced
this pull request
Jul 21, 2026
…-self-heal fix(cli): recover leaked mouse tracking terminal state
Gravezzz
pushed a commit
to Gravezzz/hermes-agent
that referenced
this pull request
Jul 21, 2026
With the tokenizer reassembling split CSI sequences across a flush (prior commit), no SGR mouse fragment can reach a text token anymore — terminals write a mouse report as one atomic sequence, and any read/flush split now re-joins in the tokenizer buffer instead of leaking. That makes the whole downstream recovery layer dead code: - SGR_MOUSE_FRAGMENT_RE, MOUSE_BURST_NOISE_RE, MOUSE_BURST_RESIDUE_RE - parseTextWithSgrMouseFragments / parseSgrMouseFragment / normalizeSgrMouseFragment - the whole-text mouse-burst noise fast path in parseMultipleKeypresses Remove all of it (~185 lines) and the tests that only exercised it. The narrow legacy X10 wheel-tail resynth stays (distinct mechanism, kept with its own test). This retires the NousResearch#17701 → NousResearch#18113 → NousResearch#26781 → NousResearch#28463 → NousResearch#35512 regex hardening chain in favor of the one correct parser fix.
leewenjie
pushed a commit
to leewenjie/hermes-agent
that referenced
this pull request
Aug 7, 2026
With the tokenizer reassembling split CSI sequences across a flush (prior commit), no SGR mouse fragment can reach a text token anymore — terminals write a mouse report as one atomic sequence, and any read/flush split now re-joins in the tokenizer buffer instead of leaking. That makes the whole downstream recovery layer dead code: - SGR_MOUSE_FRAGMENT_RE, MOUSE_BURST_NOISE_RE, MOUSE_BURST_RESIDUE_RE - parseTextWithSgrMouseFragments / parseSgrMouseFragment / normalizeSgrMouseFragment - the whole-text mouse-burst noise fast path in parseMultipleKeypresses Remove all of it (~185 lines) and the tests that only exercised it. The narrow legacy X10 wheel-tail resynth stays (distinct mechanism, kept with its own test). This retires the NousResearch#17701 → NousResearch#18113 → NousResearch#26781 → NousResearch#28463 → NousResearch#35512 regex hardening chain in favor of the one correct parser fix.
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
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.
Summary
?1006/?1003/?1002/?1000, focus, paste, alt-screen, Kitty keyboard, modifyOtherKeys) when the TUI starts and again on graceful signal/OOM shutdownESC[<...M/m), including visible and bare degraded forms, and recover terminal modes in-place if they are observedTest plan
scripts/run_tests.sh tests/cli/test_cli_terminal_response_sanitizer.pyscripts/run_tests.sh tests/cli/test_cli_bracketed_paste_sanitizer.py tests/cli/test_cli_terminal_response_sanitizer.pynpm test -- --run src/__tests__/terminalModes.test.tsnpm run type-check