Argument/flag validation failures in the minimal CLI framework threw a
plain Error that bubbled to the process-level catch in cli.ts, which
dumps a Bun.inspect code frame — a minified dist/cli.js excerpt in
compiled binaries. A missing model on `omp bench` looked like a crash.
- Add CliUsageError; throw it from Command.parse for missing/invalid
args and flags.
- Catch CliUsageError in run(): print `error: <msg>` plus the command
usage line to stderr, exit 1, no stack.
- Render required variadic positionals as MODELS... in usage, not the
misleading optional [MODELS]; extract commandUsageLine helper.
Fixes#5369
- Added bounded A* routing extents plus an expansion backstop so unreachable attachment points return null instead of searching the unbounded quadrant.
- Covered the reported declaration-order graph and an enclosed destination attachment point.
- Documented the Mermaid ASCII routing fix in the utils changelog.
Fixes#5293
- Relocated `AsyncDrain` class from `coding-agent` to `utils` package.
- Updated `HistoryStorage` to import `AsyncDrain` from shared utilities.
- Centralized the utility to allow reuse across the codebase.
On a fresh profile ~/.omp/logs may not exist yet (the logger creates it
lazily), so opening getLogPath() with "a" threw and the guard fell back to
/dev/null — discarding macOS diagnostics and native crash reports instead
of preserving them in the omp log. mkdir the redirect target's parent
(recursive) before opening; the /dev/null fallback stays as the safety net.
On macOS, libmalloc writes runtime diagnostics (e.g. "MallocStackLogging:
can't turn off malloc stack logging because it was not enabled") directly
to fd 2 of the running TUI process at arbitrary times, painting into the
viewport. The existing env-strip only protects child processes.
Add a fd-level stderr guard in pi-utils (suppressTerminalStderr /
restoreTerminalStderr) that dup2-redirects fd 2 to the omp log file while
the TUI owns the terminal, and restores it at every ownership handoff
(external editor, Ctrl+Z suspend, shutdown, crash restore). Postmortem
fatal handlers restore fd 2 before printing so crash reports stay visible.
Mirrors openai/codex#24459.
- Introduced `stringifyJson` helper to preserve bigint precision by serializing them as decimal strings.
- Replaced native `JSON.stringify` across compaction and session management modules to prevent serialization errors when handling bigint values in tool arguments.
- Added regression tests in `agent` and `coding-agent` packages to ensure bigint tool arguments remain intact through compaction and persistence flows.
Moved the config.yml/config.yaml filename order into pi-utils as
MAIN_CONFIG_FILENAMES and taught the auth-broker config reader to probe
both extensions with the same precedence as the settings loader.
Fixes#4914
- Introduced a rejection interception mechanism to capture unhandled promise rejections from eval cell code.
- Attributed floating rejections to specific runs to fail the owning cell instead of crashing the process or worker.
- Downgraded rejections occurring after a cell finished to warn logs to prevent silent failures.
Filtered Bun-autoloaded launch .env.local entries out of child shell environments so nested commands can load their own dotenv files.
Added a regression test covering Convex-style inherited deployment variables while preserving ordinary inherited env values.
Fixes#4723
- Added comprehensive test suite for `postmortem` utility error handling, covering cleanup symbol marking, cause chain depth validation, and process exception suppression.
- Added tests for `browser-run-cancellation` ensuring proper `unhandledRejection` suppression and correct `ToolAbortError` propagation during teardown.
- Implemented `collectUnhandledRejections` helper to verify silence of process-level rejections during async race conditions in browser runs.
- Added integration-style probe tests for `postmortem` to verify that marked cleanup errors allow process survival while unmarked ones remain fatal.
- Added markExpectedCleanupError and isExpectedCleanupError to identify and downgrade routine teardown errors to warnings.
- Updated global uncaughtException and unhandledRejection handlers to suppress process exit for tagged cleanup errors.
- Introduced a 10s execution deadline for cleanup callbacks to prevent process hanging during teardown.
- Mark demoted thinking blocks with a symbol to decouple paragraph separators from the raw text content.
- Update `openai-completions` to conditionally insert a newline only when a demoted thinking block is followed by other content.
- Ensure terminal demoted thinking blocks are trimmed to prevent trailing whitespace that causes rejection by the Anthropic API.
- Added scripts/fix-dts-extensions.test.ts to the test:scripts npm script.
- Included the .d.ts extension rewrite test in the CI workspace test suite.
- Updated repoScriptTests list in scripts/ci-test-ts.ts to ensure consistent tracking.
- Implemented classifyJsonPrefix to categorize buffers as complete, valid prefix, or invalid based on RFC 8259 strictness.
- Added utility to support disambiguation of identifierless streaming tool-call deltas during model response processing.
- Validated classifier logic with comprehensive suite covering nested structures, escape sequences, and strict formatting rules.
can1357