Skip to content

Exclude code-quality sweep artifacts from provider guard#203

Draft
nestharus wants to merge 51 commits into
mainfrom
cq/guard-exclude-sweep
Draft

Exclude code-quality sweep artifacts from provider guard#203
nestharus wants to merge 51 commits into
mainfrom
cq/guard-exclude-sweep

Conversation

@nestharus

Copy link
Copy Markdown
Owner

Summary:

  • Add a git pathspec exclusion for planning/code-quality-sweep/** to the AGE-244 provider-name grep guard.
  • Keep the guard rooted at the repo and covering real source/scripts; only machine-local planning artifacts are excluded.

Verification:

  • OULIPOLY_DATA_DIR=$(mktemp -d) cargo test -p oulipoly-runtime --test age244_s7b_export_replace_dispatch
  • Temporary source-file provider-name hit made the guard fail as expected, then was reverted.
  • cargo fmt --check
  • cargo clippy -p oulipoly-runtime --all-targets -- -D warnings

nestharus and others added 30 commits June 26, 2026 16:18
Interactive --resume built its target from the model-effective provider,
injecting -m/--variant launch args; opencode rejects --variant and exits 1.
Use the runtime provider resume argv (session already carries model/variant).
Verified against real ses_1012 (enters TUI vs exit-1). Token-clean.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…merge

When multiple segments of one chain collapse to one post-alias tuple, select
survivor/start/tail independently: the open segment wins survivor selection,
earliest started_at and latest tail are preserved, and merged_ended_at stays
open if any merged row is open. forward.sql gains a fail-closed guard against
plans that would close an open merge group. Synthetic-fixture tests only.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
main had accumulated rustfmt drift across ~24 files; every pipeline worktree
regenerated these on 'cargo fmt', polluting WU diffs and causing repeated
'rogue edits' confusion. Pure formatting, no behavior change.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
Live pty_interactive owners that NACK immediate notify-time PTY delivery
(e.g. unsafe_mid_line) left a durable pending mailbox row with no wake claim,
and sweep filtered busy sessions out before any retry -> the owner showed a
pending count but was never notified (affects opencode + claude interactive
sessions). Add a sweep pre-pass that retries pending rows via the PTY control
socket for live pty_interactive owners, marking delivered only after a PTY ACK;
no headless wake claim for busy owners; repeated NACK stays safely pending.
Sidecar-only pending lookup; no schema change; no provider-specific tokens.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…patches

Use 'agent-bash run -- agents ...' (not raw agents) so caller ancestry is
captured and agent_bash_complete is delivered via the mailbox/wake system.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…ence

Make default-provider REPL sessions visible to invocation/session-chain
machinery (persist reusable cwd evidence), and add a provider-neutral
script-storage cwd fallback. Resume precedence corrected so live external-provider
session_runtime.effective_cwd wins, with documented fallback to recorded
account cwd then cwd_script. Provider-neutral (claude token delta -4); synthetic
tests cover both stale and live cases. (supersedes draft PR #202)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
Provider-ref Claude sessions dispatch through the provider seam; legacy/builtin
sessions keep the in-tree reader path; source guard asserts builtin readers stay
present (deletion blocked without updating the proof). Test-only; merged with
session-cwd test imports.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
… None (squash)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
nestharus and others added 21 commits June 28, 2026 07:34
…atch (SI WU-1)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…t, always-visible input, filterable list (default bash), click-toggle-close, bar closes list, remove Ctrl+O (E1)

Bottom-pane input box now inserts a newline on Ctrl+Enter (ctrl_enter_sequence_len)
and always renders without manual expand; the process list is filterable by mode
(MonitorViewMode label: bash/procs/mailbox, default bash), clicking an open item
closes it, clicking the overlay bar closes the list, and the Ctrl+O focus-toggle
byte is removed. Multi-line pseudo-input cursor (pseudo_input_cursor_line_col).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…Ds [SI WU-5]

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…ts at cursor (E3)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…cident fix)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…sion import [SI WU-8]

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…encode [SI WU-9]

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…on providers [SI WU-10]

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
The bottom pseudo-input box mapped BOTH CR (0x0d) and LF (0x0a) to submit. Terminals
commonly send CR for Enter and LF (Ctrl+J) for Ctrl+Enter, so Ctrl+Enter was submitting
instead of inserting a newline (confirmed via overlay-input-trace.log: Ctrl+Enter -> 0a).
Now CR submits and LF inserts a newline at the cursor (also makes a pasted raw '\n'
insert rather than submit). CSI-u Ctrl+Enter still handled by ctrl_enter_sequence_len.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…input above OBS (E5)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…ver, schemas) [cq-fix1]

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…rs [cq-config-fix1]

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…ad/store [cq-agentcrates-fix1]

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…epositories [cq-config-fix2]

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…ation/working_set [cq]

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
…/discovery [cq]

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K7cQsdbQh7wM1dRRjyjYn5
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.

1 participant