fix(windows): capture is not a no-window boundary; route flashing spawns through chokepoint - #53829
Merged
Merged
Conversation
Contributor
🔎 Lint report:
|
| Rule | Count |
|---|---|
unresolved-attribute |
2 |
First entries
run_agent.py:3014: [unresolved-attribute] unresolved-attribute: Object of type `Self@get_credits_spent_micros` has no attribute `_credits_session_start_micros`
tests/run_agent/test_credits_notices_toggle.py:76: [unresolved-attribute] unresolved-attribute: Unresolved attribute `_credits_session_start_micros` on type `AIAgent`
✅ Fixed issues (1):
| Rule | Count |
|---|---|
invalid-assignment |
1 |
First entries
tests/run_agent/test_credits_notices_toggle.py:76: [invalid-assignment] invalid-assignment: Object of type `None` is not assignable to attribute `_credits_session_start_micros` of type `int`
Unchanged: 6070 pre-existing issues carried over.
Diagnostics are surfaced as warnings — this check never fails the build.
…wns through chokepoint Follow-up to #53791 addressing review feedback: the footgun checker treated capture_output=/stdout=/stderr=/check_output as proof a subprocess can't pop a Windows console. That invariant is false — stream redirection controls where a child's output goes, not whether a console is allocated. From a console-less parent (Desktop/Electron, pythonw.exe, detached gateway/cron) a console-subsystem child still flashes a window even when fully captured. - check-windows-footguns.py: capture/redirect/check_output is no longer a blanket safe-pass. Added _WINDOWS_FLASHING_PROGRAMS (git/gh/npm/node/python/uv/ffmpeg/ docker/powershell/…); calls to those are flagged even when captured. Non-flashing programs keep the capture exemption (no 271-site noise). _subprocess_compat.run/ popen calls are inherently safe (wrapper injects CREATE_NO_WINDOW). - Routed the 35 genuine flashing git/gh/npm/uv/ffmpeg/docker spawns through the _subprocess_compat.run/popen chokepoint (Brooklyn's wrapper from #53810) — the durable fix, not per-site annotations. cmd.exe /c start stays # ok (intentional). - Updated tests + CONTRIBUTING.md rule #17 to the corrected invariant.
teknium1
force-pushed
the
fix/footgun-capture-invariant
branch
from
June 27, 2026 21:20
044b863 to
6177a62
Compare
alt-glitch
added
type/bug
This was referenced Jun 27, 2026
Closed
This was referenced Jun 27, 2026
teknium1
added a commit
that referenced
this pull request
Jun 27, 2026
…53853) * Revert "fix(windows): capture is not a no-window boundary; route flashing spawns through chokepoint (#53829)" This reverts commit 2ecca1e. * Revert "fix(windows): stop terminal-window popups from background spawns (#53810)" This reverts commit 5db1430. * Revert "fix(windows): stop subprocess console-window popups + add CI guard (#53791)" This reverts commit ef17cd2.
19 tasks
1 task
lpaiu-cs
added a commit
to lpaiu-cs/hermes-agent
that referenced
this pull request
Jun 28, 2026
…ch#53853 revert The first cut routed these probes through _subprocess_compat.run/popen (the NousResearch#53810 chokepoint), but NousResearch#53853 reverted NousResearch#53791/NousResearch#53810/NousResearch#53829 — so those calls would now AttributeError on main (the wrappers are gone; windows_hide_flags() survived). Switch the two sites to pass creationflags=windows_hide_flags() directly, the same surviving-helper approach NousResearch#53892 took post-revert. Also drop the now-moot footgun changes: the console-spawn rule was reverted, so adding tui_gateway to its --all roots no longer buys anything (and would only surface unrelated, hasattr-guarded signal handlers in entry.py). Net change is now just the two gateway probe sites + the test, which asserts the no-window flag is passed. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
pai-scaffolde
pushed a commit
to pai-scaffolde/hermes-agent
that referenced
this pull request
Jun 28, 2026
…wns through chokepoint (NousResearch#53829) Follow-up to NousResearch#53791 addressing review feedback: the footgun checker treated capture_output=/stdout=/stderr=/check_output as proof a subprocess can't pop a Windows console. That invariant is false — stream redirection controls where a child's output goes, not whether a console is allocated. From a console-less parent (Desktop/Electron, pythonw.exe, detached gateway/cron) a console-subsystem child still flashes a window even when fully captured. - check-windows-footguns.py: capture/redirect/check_output is no longer a blanket safe-pass. Added _WINDOWS_FLASHING_PROGRAMS (git/gh/npm/node/python/uv/ffmpeg/ docker/powershell/…); calls to those are flagged even when captured. Non-flashing programs keep the capture exemption (no 271-site noise). _subprocess_compat.run/ popen calls are inherently safe (wrapper injects CREATE_NO_WINDOW). - Routed the 35 genuine flashing git/gh/npm/uv/ffmpeg/docker spawns through the _subprocess_compat.run/popen chokepoint (Brooklyn's wrapper from NousResearch#53810) — the durable fix, not per-site annotations. cmd.exe /c start stays # ok (intentional). - Updated tests + CONTRIBUTING.md rule #17 to the corrected invariant.
pai-scaffolde
pushed a commit
to pai-scaffolde/hermes-agent
that referenced
this pull request
Jun 28, 2026
…Research#53810 NousResearch#53829 (NousResearch#53853) * Revert "fix(windows): capture is not a no-window boundary; route flashing spawns through chokepoint (NousResearch#53829)" This reverts commit 2ecca1e. * Revert "fix(windows): stop terminal-window popups from background spawns (NousResearch#53810)" This reverts commit 5db1430. * Revert "fix(windows): stop subprocess console-window popups + add CI guard (NousResearch#53791)" This reverts commit ef17cd2.
waefrebeorn
pushed a commit
to waefrebeorn/slermes
that referenced
this pull request
Jul 2, 2026
…wns through chokepoint (NousResearch#53829) Follow-up to NousResearch#53791 addressing review feedback: the footgun checker treated capture_output=/stdout=/stderr=/check_output as proof a subprocess can't pop a Windows console. That invariant is false — stream redirection controls where a child's output goes, not whether a console is allocated. From a console-less parent (Desktop/Electron, pythonw.exe, detached gateway/cron) a console-subsystem child still flashes a window even when fully captured. - check-windows-footguns.py: capture/redirect/check_output is no longer a blanket safe-pass. Added _WINDOWS_FLASHING_PROGRAMS (git/gh/npm/node/python/uv/ffmpeg/ docker/powershell/…); calls to those are flagged even when captured. Non-flashing programs keep the capture exemption (no 271-site noise). _subprocess_compat.run/ popen calls are inherently safe (wrapper injects CREATE_NO_WINDOW). - Routed the 35 genuine flashing git/gh/npm/uv/ffmpeg/docker spawns through the _subprocess_compat.run/popen chokepoint (Brooklyn's wrapper from NousResearch#53810) — the durable fix, not per-site annotations. cmd.exe /c start stays # ok (intentional). - Updated tests + CONTRIBUTING.md rule NousResearch#17 to the corrected invariant.
waefrebeorn
pushed a commit
to waefrebeorn/slermes
that referenced
this pull request
Jul 2, 2026
…Research#53810 NousResearch#53829 (NousResearch#53853) * Revert "fix(windows): capture is not a no-window boundary; route flashing spawns through chokepoint (NousResearch#53829)" This reverts commit bc1c8f2. * Revert "fix(windows): stop terminal-window popups from background spawns (NousResearch#53810)" This reverts commit 40591a4. * Revert "fix(windows): stop subprocess console-window popups + add CI guard (NousResearch#53791)" This reverts commit c2eab69.
Jasper6439
pushed a commit
to Jasper6439/hermes-agent
that referenced
this pull request
Jul 5, 2026
…wns through chokepoint (NousResearch#53829) Follow-up to NousResearch#53791 addressing review feedback: the footgun checker treated capture_output=/stdout=/stderr=/check_output as proof a subprocess can't pop a Windows console. That invariant is false — stream redirection controls where a child's output goes, not whether a console is allocated. From a console-less parent (Desktop/Electron, pythonw.exe, detached gateway/cron) a console-subsystem child still flashes a window even when fully captured. - check-windows-footguns.py: capture/redirect/check_output is no longer a blanket safe-pass. Added _WINDOWS_FLASHING_PROGRAMS (git/gh/npm/node/python/uv/ffmpeg/ docker/powershell/…); calls to those are flagged even when captured. Non-flashing programs keep the capture exemption (no 271-site noise). _subprocess_compat.run/ popen calls are inherently safe (wrapper injects CREATE_NO_WINDOW). - Routed the 35 genuine flashing git/gh/npm/uv/ffmpeg/docker spawns through the _subprocess_compat.run/popen chokepoint (Brooklyn's wrapper from NousResearch#53810) — the durable fix, not per-site annotations. cmd.exe /c start stays # ok (intentional). - Updated tests + CONTRIBUTING.md rule NousResearch#17 to the corrected invariant.
Jasper6439
pushed a commit
to Jasper6439/hermes-agent
that referenced
this pull request
Jul 5, 2026
…Research#53810 NousResearch#53829 (NousResearch#53853) * Revert "fix(windows): capture is not a no-window boundary; route flashing spawns through chokepoint (NousResearch#53829)" This reverts commit 2ecca1e. * Revert "fix(windows): stop terminal-window popups from background spawns (NousResearch#53810)" This reverts commit 5db1430. * Revert "fix(windows): stop subprocess console-window popups + add CI guard (NousResearch#53791)" This reverts commit ef17cd2.
habarmc1223-sudo
pushed a commit
to habarmc1223-sudo/hermes-agent-fluxmem
that referenced
this pull request
Jul 8, 2026
…wns through chokepoint (NousResearch#53829) Follow-up to NousResearch#53791 addressing review feedback: the footgun checker treated capture_output=/stdout=/stderr=/check_output as proof a subprocess can't pop a Windows console. That invariant is false — stream redirection controls where a child's output goes, not whether a console is allocated. From a console-less parent (Desktop/Electron, pythonw.exe, detached gateway/cron) a console-subsystem child still flashes a window even when fully captured. - check-windows-footguns.py: capture/redirect/check_output is no longer a blanket safe-pass. Added _WINDOWS_FLASHING_PROGRAMS (git/gh/npm/node/python/uv/ffmpeg/ docker/powershell/…); calls to those are flagged even when captured. Non-flashing programs keep the capture exemption (no 271-site noise). _subprocess_compat.run/ popen calls are inherently safe (wrapper injects CREATE_NO_WINDOW). - Routed the 35 genuine flashing git/gh/npm/uv/ffmpeg/docker spawns through the _subprocess_compat.run/popen chokepoint (Brooklyn's wrapper from NousResearch#53810) — the durable fix, not per-site annotations. cmd.exe /c start stays # ok (intentional). - Updated tests + CONTRIBUTING.md rule NousResearch#17 to the corrected invariant.
habarmc1223-sudo
pushed a commit
to habarmc1223-sudo/hermes-agent-fluxmem
that referenced
this pull request
Jul 8, 2026
…Research#53810 NousResearch#53829 (NousResearch#53853) * Revert "fix(windows): capture is not a no-window boundary; route flashing spawns through chokepoint (NousResearch#53829)" This reverts commit 710ac05. * Revert "fix(windows): stop terminal-window popups from background spawns (NousResearch#53810)" This reverts commit 3db485a. * Revert "fix(windows): stop subprocess console-window popups + add CI guard (NousResearch#53791)" This reverts commit 80b090d.
santhreal
pushed a commit
to santhreal/hermes-agent
that referenced
this pull request
Jul 13, 2026
…wns through chokepoint (NousResearch#53829) Follow-up to NousResearch#53791 addressing review feedback: the footgun checker treated capture_output=/stdout=/stderr=/check_output as proof a subprocess can't pop a Windows console. That invariant is false — stream redirection controls where a child's output goes, not whether a console is allocated. From a console-less parent (Desktop/Electron, pythonw.exe, detached gateway/cron) a console-subsystem child still flashes a window even when fully captured. - check-windows-footguns.py: capture/redirect/check_output is no longer a blanket safe-pass. Added _WINDOWS_FLASHING_PROGRAMS (git/gh/npm/node/python/uv/ffmpeg/ docker/powershell/…); calls to those are flagged even when captured. Non-flashing programs keep the capture exemption (no 271-site noise). _subprocess_compat.run/ popen calls are inherently safe (wrapper injects CREATE_NO_WINDOW). - Routed the 35 genuine flashing git/gh/npm/uv/ffmpeg/docker spawns through the _subprocess_compat.run/popen chokepoint (Brooklyn's wrapper from NousResearch#53810) — the durable fix, not per-site annotations. cmd.exe /c start stays # ok (intentional). - Updated tests + CONTRIBUTING.md rule NousResearch#17 to the corrected invariant.
santhreal
pushed a commit
to santhreal/hermes-agent
that referenced
this pull request
Jul 13, 2026
…Research#53810 NousResearch#53829 (NousResearch#53853) * Revert "fix(windows): capture is not a no-window boundary; route flashing spawns through chokepoint (NousResearch#53829)" This reverts commit 12aa1c3. * Revert "fix(windows): stop terminal-window popups from background spawns (NousResearch#53810)" This reverts commit ec5d0c1. * Revert "fix(windows): stop subprocess console-window popups + add CI guard (NousResearch#53791)" This reverts commit d0ffccd.
Gravezzz
pushed a commit
to Gravezzz/hermes-agent
that referenced
this pull request
Jul 21, 2026
…wns through chokepoint (NousResearch#53829) Follow-up to NousResearch#53791 addressing review feedback: the footgun checker treated capture_output=/stdout=/stderr=/check_output as proof a subprocess can't pop a Windows console. That invariant is false — stream redirection controls where a child's output goes, not whether a console is allocated. From a console-less parent (Desktop/Electron, pythonw.exe, detached gateway/cron) a console-subsystem child still flashes a window even when fully captured. - check-windows-footguns.py: capture/redirect/check_output is no longer a blanket safe-pass. Added _WINDOWS_FLASHING_PROGRAMS (git/gh/npm/node/python/uv/ffmpeg/ docker/powershell/…); calls to those are flagged even when captured. Non-flashing programs keep the capture exemption (no 271-site noise). _subprocess_compat.run/ popen calls are inherently safe (wrapper injects CREATE_NO_WINDOW). - Routed the 35 genuine flashing git/gh/npm/uv/ffmpeg/docker spawns through the _subprocess_compat.run/popen chokepoint (Brooklyn's wrapper from NousResearch#53810) — the durable fix, not per-site annotations. cmd.exe /c start stays # ok (intentional). - Updated tests + CONTRIBUTING.md rule NousResearch#17 to the corrected invariant.
Gravezzz
pushed a commit
to Gravezzz/hermes-agent
that referenced
this pull request
Jul 21, 2026
…Research#53810 NousResearch#53829 (NousResearch#53853) * Revert "fix(windows): capture is not a no-window boundary; route flashing spawns through chokepoint (NousResearch#53829)" This reverts commit 2ecca1e. * Revert "fix(windows): stop terminal-window popups from background spawns (NousResearch#53810)" This reverts commit 5db1430. * Revert "fix(windows): stop subprocess console-window popups + add CI guard (NousResearch#53791)" This reverts commit ef17cd2.
leewenjie
pushed a commit
to leewenjie/hermes-agent
that referenced
this pull request
Aug 7, 2026
…wns through chokepoint (NousResearch#53829) Follow-up to NousResearch#53791 addressing review feedback: the footgun checker treated capture_output=/stdout=/stderr=/check_output as proof a subprocess can't pop a Windows console. That invariant is false — stream redirection controls where a child's output goes, not whether a console is allocated. From a console-less parent (Desktop/Electron, pythonw.exe, detached gateway/cron) a console-subsystem child still flashes a window even when fully captured. - check-windows-footguns.py: capture/redirect/check_output is no longer a blanket safe-pass. Added _WINDOWS_FLASHING_PROGRAMS (git/gh/npm/node/python/uv/ffmpeg/ docker/powershell/…); calls to those are flagged even when captured. Non-flashing programs keep the capture exemption (no 271-site noise). _subprocess_compat.run/ popen calls are inherently safe (wrapper injects CREATE_NO_WINDOW). - Routed the 35 genuine flashing git/gh/npm/uv/ffmpeg/docker spawns through the _subprocess_compat.run/popen chokepoint (Brooklyn's wrapper from NousResearch#53810) — the durable fix, not per-site annotations. cmd.exe /c start stays # ok (intentional). - Updated tests + CONTRIBUTING.md rule NousResearch#17 to the corrected invariant.
leewenjie
pushed a commit
to leewenjie/hermes-agent
that referenced
this pull request
Aug 7, 2026
…Research#53810 NousResearch#53829 (NousResearch#53853) * Revert "fix(windows): capture is not a no-window boundary; route flashing spawns through chokepoint (NousResearch#53829)" This reverts commit 2ecca1e. * Revert "fix(windows): stop terminal-window popups from background spawns (NousResearch#53810)" This reverts commit 5db1430. * Revert "fix(windows): stop subprocess console-window popups + add CI guard (NousResearch#53791)" This reverts commit ef17cd2.
Oc2cO
pushed a commit
to Oc2cO/hermes-agent
that referenced
this pull request
Aug 16, 2026
…revert d3d621f Root cause of the persisting desktop console flash: commit d3d621f bulk- reverted three earlier windows-hide fix PRs (NousResearch#53791/NousResearch#53810/NousResearch#53829) across ~40 files in June; only a handful were individually re-fixed afterward. This restores creationflags=windows_hide_flags() (or windows_detach_popen_kwargs() for the one long-running Popen in raft/adapter.py) on every remaining Windows-reachable subprocess/asyncio spawn found by a full-repo audit: cli.py (24 sites: worktree/branch-resolution helpers on the `hermes -w` startup path), kanban_db.py (6, gateway boot/orphan-reconcile), gateway/run.py (2, asyncio create_subprocess_exec/shell), clipboard.py (4), claw.py (2), mcp_catalog.py, dep_ensure.py, doctor.py, setup.py, dump.py, webhook.py, raft/adapter.py, managed_uv.py (11), mem0/_setup.py (7), honcho/client.py, iron_proxy.py (7), environments/docker.py (19), environments/ssh.py (11), uninstall.py, plugins_cmd.py, profiles.py, secret_sources/{bitwarden,onepassword}.py, discord/voice_mixer.py, contributor_audit.py, profile_distribution.py. Deliberately left visible (interactive/foreground by design): editor launches, `claude setup-token` OAuth login, relaunch.py's self-relaunch, and anything gated to non-Windows-only code paths (service_manager.py's s6-overlay branch, command.py's secret helper, profile-tui.py's pty import). Verified via ast.parse + py_compile on every changed file and a live import check on all touched non-CLI modules. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
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
2 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.
Summary
Follow-up to #53791 correcting the false invariant Gille flagged in review: the Windows footgun checker treated
capture_output=/stdout=/stderr=/check_outputas proof a subprocess can't pop a console. That's wrong — stream redirection controls where a child's output goes, not whether a console is allocated. From a console-less parent (Desktop/Electron,pythonw.exe, a detached gateway/cron) a console-subsystem child still flashes a window even when fully captured.Changes
scripts/check-windows-footguns.py— capture/redirect/check_outputis no longer a blanket safe-pass. Added_WINDOWS_FLASHING_PROGRAMS(git, gh, npm, node, python, uv, ffmpeg, docker, powershell, …); calls to those are flagged even when captured. Non-flashing programs keep the capture exemption, so the rule stays comprehensive without flipping ~271 POSIX-only/dynamic calls red. Calls routed through_subprocess_compat.run/popenare inherently safe (the wrapper carries the flag)._subprocess_compat.run/popenchokepoint added in fix(windows): stop terminal-window popups from background spawns #53810 — the durable fix, not per-site annotations.cmd.exe /c startstays# windows-footgun: ok(intentional window).test_flags_flashing_programs_even_when_captured), plusCONTRIBUTING.mdrule Add support for Atropos Agentic RL environments (requires branch tool… #17 rewritten.Validation
git/npmspawn from GUI parentcheck-windows-footguns.py --alltests/agent/regressionBuilds on the wrapper chokepoint from #53810: "no visible terminal" is structural for these spawns, not a per-call-site rule that gets forgotten.
Infographic