Compare commits

..

8 commits

Author SHA1 Message Date
andrewgu-oai
f2908e170c [0.159] Backport account security setup reminders for 0.159.3 (#49744)
Backport #49715 to the 0.159 maintenance line for 0.159.3 so local
ChatGPT sessions can display optional account-security setup reminders.

This is an unchanged cherry-pick of the upstream patch: 23 files,
including authenticated-request, stale-account, notice-validation and
TUI regression tests. The server continues to own eligibility and
rollout; unavailable notices produce no banner.

Validation: clean cherry-pick onto release/0.159 at
2851358eae; stable patch ID matches
upstream; whitespace checks pass. Exact-backport CI and production build
validation are pending.

Co-authored-by: dennyku <dennyku@openai.com>
2026-10-02 19:04:40 +00:00
andrewgu-oai
94cd1062e9 [0.159] Backport Windows console suppression for 0.159.2 (#49385)
Backport #49164 and #49308 onto the content commit of 0.159.1 to
suppress Windows console flashing reported in #48074. The two
cherry-picks applied without conflicts; no other fixes are included.
Cargo remains at the release-branch content version, 0.0.0.

Validation: `git diff --check` passes. Native Windows execution has not
been validated on this exact candidate. Keep this PR in draft pending
required review and a focused Windows smoke test: from a non-elevated
PowerShell host with workspace-write and the unelevated sandbox, verify
built-ins and external whoami.exe, cmd.exe, Git and nested PowerShell
terminate with expected output, both with a matching managed daemon and
with --no-daemon, without visible console flashes. Record exact
binary/candidate SHA, CLI and daemon versions, CODEX_HOME,
configuration, exit codes and durations.

The alpha.2 reports in
https://github.com/openai/codex/issues/48074#issuecomment-5898123863 and
https://github.com/openai/codex/issues/48074#issuecomment-5897958690
conflict on execution safety; they do not establish a causal regression
or validate this exact backport. Do not publish 0.159.2 until the
evidence and required checks/review are complete.

---------

Co-authored-by: Eric Traut <etraut@openai.com>
2026-10-02 19:04:39 +00:00
Celia Chen
c46b6c7048 [0.159] Backport GPT-6.1 Sol Bedrock catalogs (#49342)
Backport #49339 onto `release/0.159` for 0.159.1. This adds GPT-6.1 Sol
to the Bedrock Mantle and Runtime catalogs and makes it the default
while retaining existing models and provider capability normalization.

The bundled catalog dependency is already on the release branch through
#49323. The cherry-pick applied without conflicts and has the same
stable patch ID as the original public commit.

Validation:
- `just test -p codex-model-provider -p codex-model-provider-info --lib
bedrock`: 57 passed.
- `just test -p codex-tui --lib bedrock`: 7 passed; reviewed Mantle and
Runtime model-picker snapshots, including GovCloud coverage.
- `just test -p codex-app-server --test all -E
'test(suite::v2::thread_start::thread_start_provider_model_fallback_) or
test(suite::v2::thread_start::thread_start_bedrock_runtime_)'`: 4 passed
on automatic retry. All four initial attempts timed out during
initialization; nextest reports these as flaky passes.
- `just fmt` completed successfully. Discarded unrelated justfile-only
formatting churn; the backport remains identical to the public source
patch.
- `git diff --check`: passed.
- Live AWS inference has not been verified as part of this backport.

CI remains blocked by failures outside this nine-file backport:
- [Manifest
check](https://github.com/openai/codex/actions/runs/36615506389/job/109567359534):
stale `code-mode` feature exception.
-
[Codespell](https://github.com/openai/codex/actions/runs/36615506389/job/109567359001):
`WRONLY` and `oint` in unchanged exec-server/TUI math files.
- [cargo
shear](https://github.com/openai/codex/actions/runs/36615506389/job/109567775485):
unlinked `core/tests/suite/scenarios_read_only_mcp.rs` and
`login/src/internal_identity.rs`.
- [Bazel
Clippy](https://github.com/openai/codex/actions/runs/36615506389/job/109567360047):
unused imports in unchanged core implementation/test files.
- [SDK
checks](https://github.com/openai/codex/actions/runs/36615506389/job/109567359667):
existing formatter-driver test expects direct rustfmt while the driver
uses cargo fmt; neither file changes here.

Release coordination: include this backport together with #49323 before
publishing 0.159.1. Publication remains with the release captain.
2026-10-02 19:04:39 +00:00
andrewgu-oai
2afbde2f7d [0.159] Prepare 0.159.1 release backports (#49323)
Prepare the backports for 0.159.1 on `release/0.159`. The release set is
incomplete: one additional backport is pending. Keep this PR in draft
until the full release set is included.

Currently includes #49318: add GPT-6.1 Sol as the default bundled model
and update the older Sol descriptions and model-picker snapshots.

The backport omits the `skill_catalog_dedup` snapshot because its test
does not exist on the 0.159 release line. The model catalog and all
retained snapshot files match the original public commit.

Release-line CI currently reports a stale `code-mode` feature exception
in the manifest checker, unlinked files in `codex-core` and
`codex-login`, and Codespell findings in the exec-server and TUI math
code. The reported files are unchanged from the 0.159.0 content commit.
A Linux Bazel job also failed while downloading Buildozer with HTTP 500.
2026-10-02 19:04:39 +00:00
eknight-oai
00217c376d Prevent message-board notifications from reopening final answers (#48982)
## Why

Message-board notifications delivered as ordinary injected input could cause an agent to sample again after its final answer.

## What changed

Route notifications through the current turn's mailbox as `InterAgentCommunication`, checking atomically that the turn is running and still accepts mailbox delivery. Skip notifications when delivery is closed. Accepted notifications may remain in conversation history for a later turn without reopening the finalized answer or starting a new turn.

Report pending inter-agent communications as mailbox activity, while preserving priority for steering input.

## Testing

Add regression scenarios for notifications arriving before and after finalization, asserting no extra sampling and the expected notification presence in the next turn. Extend input-queue coverage to recognize already-pending mailbox activity.

GitOrigin-RevId: 41d786041b4eccc5ccf926f2f190473676a3ac2f
2026-10-02 19:04:39 +00:00
jif
df32b8dadf Isolate realtime auth fallback test from startup prewarm (#48973)
## Why

Startup prewarm can reconnect after its empty WebSocket connection closes,
consuming the scripted realtime session reply in the environment-key fallback
test with ChatGPT authentication.

## What changed

Use separate WebSocket servers for startup and realtime traffic, configuring
`experimental_realtime_ws_base_url` to point to the realtime server. Assert that
it receives exactly one handshake and that the handshake uses the environment
key for authorization.

GitOrigin-RevId: 0e8ba4bdd66ef8aa3c619e36454dded9dcf55444
2026-10-02 19:04:39 +00:00
Eric Traut
6e4b58112d Expand native Mermaid flowchart syntax support (#48895)
## What changed

- Default `flowchart` and `graph` to top-down layout when no direction is specified.
- Support solid and dashed edges with directed, undirected, or bidirectional endpoints: `-->`, `---`, `<-->`, `-.->`, `-.-`, and `<-.->`.
- Accept pipe labels on these edges and spaced directed infix labels such as `A -- send --> B` and `A -. retry .-> B`.
- Expand `&` groups into all source-to-target edges, including in chains. For example, `A & B --> C & D` creates four edges. Enforce the 24-edge limit after expansion and cap each group at 24 node references.

## Testing

Add coverage for equivalent syntax, edge styles and endpoints, malformed input, and expansion limits. Extend rendering snapshots across all four directions and the TUI native-renderer test with the new syntax.

GitOrigin-RevId: 6e51653ebf6b8b7e7e1edec0c2e09a5d9c8dec74
2026-10-02 19:04:39 +00:00
andrewgu-oai
51215e9f88
Backport MXC PowerShell fix and safe alpha publishing to 0.159.0-alpha.12 (#50061)
Backport two reviewed upstream fixes onto the exact content parent of
0.159.0-alpha.12 for the desktop 260930 train:

- Public #49019 (internal #10589),
4fd5745e84: reuse unpackaged PowerShell
fallback for local MXC. Remote, unsandboxed, restricted-token handling
and elevated-only -NoProfile behavior are preserved.
- Public #49704, 5602705c96: keep newer
npm alpha pointers unchanged when publishing an older frozen hotfix. The
existing OIDC publisher uses version-specific tags; registry lookup
failure aborts publication. No new credentials or permission changes.

Both patches apply cleanly and match their upstream patch IDs. Eight
files total, +360/-71; no extra dependencies or unrelated product
changes. Frozen base: 1cc7e23612. Intended
publication: 0.159.0-alpha.12.1 through the exact-SHA-guarded branch
publisher.

Validation: just fmt and whitespace checks passed; all 50 focused
PowerShell library tests passed; all 17 release/tag-selection tests
passed. Initial product-only head passed all three production build
checks (Linux/macOS/Windows); updated-head checks are pending. Native
MXC execution is not claimed by Linux validation. The included Windows
integration test skips if MXC is unavailable.

Initial CI manifest-policy failure reproduces on the exact frozen base;
Codespell and cargo-shear failures concern unchanged files. No unrelated
lint fixes included.

---------

Co-authored-by: Ian MacLeod <imac@openai.com>
2026-10-01 11:40:51 -07:00
3608 changed files with 61627 additions and 209899 deletions

Can't render this file because it has a wrong number of fields in line 9.

907
MODULE.bazel.lock generated

File diff suppressed because one or more lines are too long

Some files were not shown because too many files have changed in this diff Show more