[AlexTranAmz]

Fixes: N/A

What was the problem/requirement? (What/Why)

The Rust session runtime tags every record it emits on the openjd.sessions logger with session_id, openjd_log_content (a LogContent bitmask) and openjd_timestamp_usec. Those records reach Python through pyo3-log, which renders each key-value with its Rust Display before passing it to Logger.makeRecord(..., extra=...). So openjd_log_content arrives as a decimal string such as "8" rather than the LogContent this logger's docstring promises, and openjd_timestamp_usec as a string. Consumers route these records by type-checking openjd_log_content against openjd.sessions.LogContent (the Deadline Cloud Worker Agent does isinstance(record.openjd_log_content, LogContent)), so every Rust-emitted record fails the check and is dropped from the worker log.

This only becomes observable with an openjd-model build that enables pyo3-log's kv feature (companion PR in openjd-model-for-python); today the fields are missing entirely.

A related trap: openjd/sessions/_v1/_logging.py declared its own LogContent(Flag) with the same members and bit values as openjd/sessions/_logging.py. enum.Flag comparisons are class-scoped, so openjd.sessions._v1.LogContent.BANNER failed the same isinstance check against the public class. Nothing in _v1 tags records with that copy today, but the name is re-exported.

What was the solution? (How)

openjd/sessions/_v1/_logging.py:

• Importing openjd.sessions._v1 installs one logging.Filter on the openjd.sessions logger. If openjd_log_content is a digit string it becomes LogContent(int(value)); if openjd_timestamp_usec is a digit string it becomes an int. Anything else (absent, already typed, a bit this LogContent does not define, malformed) is left as it arrived. The filter never drops a record.
• It is a logger filter, not a handler filter: pyo3-log delivers records with Logger.handle(), which applies the logger's filters before any handler or handler-level filter sees the record, which is where consumers type-check.
• Installation is idempotent (repeated imports do not stack filters).
• LogContent is now re-exported from openjd.sessions._logging instead of redefined, so openjd.sessions._v1.LogContent is openjd.sessions.LogContent.

What is the impact of this change?

Consumers attaching handlers to openjd.sessions receive the Rust runtime's records with the documented types once an openjd-model release forwards the fields. With the current openjd-model 0.14.0 the fields are absent and the filter is a no-op. Either release order is safe: this package first is inert; openjd-model first restores session_id (correct routing) while openjd_log_content stays a string that typed consumers reject exactly as today.

How was this change tested?

See DEVELOPMENT.md for information on running tests.

• Have you run the unit tests? Yes.

• test/openjd/sessions_v1/test_logging.py: filter unit tests with logging.makeLogRecord (each LogContent member and a composite from strings; already-typed, absent, malformed and undefined-bit values preserved; timestamp conversion; session_id untouched; idempotent install; records passed through Logger.handle() reach handlers and handler-level filters normalized; _v1.LogContent is openjd.sessions.LogContent with unchanged bit values).

• test/openjd/sessions_v1/test_rust_log_tags.py: runs a real openjd.sessions._v1.Session task (sys.executable -c, so Linux/macOS/Windows) and asserts the captured records carry a LogContent, this session's session_id, an int timestamp, and that the task's stdout is tagged COMMAND_OUTPUT. The four assertions that need the fields are gated on a probe (any record carrying session_id) so they skip on openjd-model 0.14.0 and run automatically on a build that forwards the fields; the probe has its own ungated tests.

Local results (Linux, Python 3.12):

• hatch run test with openjd-model 0.14.0: 1046 passed, 50 skipped, 16 xfailed (gated tests skip as designed)
• hatch run test with a wheel built from the companion openjd-model branch: 1050 passed, 46 skipped, 16 xfailed (gated tests run and pass)
• hatch run lint (ruff, black, mypy): clean
• Not run locally: Python 3.9-3.11, 3.13, 3.14, Windows, macOS (CI matrix); cross-user suites (not touched).

Was this change documented?

• Are relevant docstrings in the code base updated? Yes: the LOG docstring documents the three attributes and the normalization; the filter and install helper carry docstrings.

Is this a breaking change?

No. openjd.sessions._v1 is private and exports the same names. openjd.sessions._v1.LogContent is now the same class object as openjd.sessions.LogContent; member names and bit values are unchanged. Cross-class comparisons that previously returned False or raised TypeError now behave as same-class comparisons, which is the intended fix.

Does this change impact security?

• Does the change need to be threat modeled? No. Rust records already classified as EXCEPTION_INFO, PROCESS_CONTROL or HOST_INFO become visible to consumers that allowlist those categories, which is the same routing the Python runtime's records have always had. No new data is produced.

Cross-port to openjd-rs

This package is being migrated to Rust in openjd-rs/crates/openjd-sessions.
Behavioral changes made here should be replicated there to keep the two
implementations in sync until the migration is complete.

• This change does not affect runtime behavior (docs / tests / tooling only), or
• A matching change has been opened in openjd-rs (link the PR here): , or
• A tracking issue has been filed in openjd-rs to port this change (link here): , or
• Cross-porting is not applicable for this change because: it is Python-side normalization of values that pyo3-log stringifies on the way into Python; openjd-rs already emits these fields as typed key-values (session_log! sends openjd_log_content as a u32 and openjd_timestamp_usec as a u64).

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[jericht]

aws-deadline/deadline-cloud-worker-agent#1130 -> will need to fix the actions here too before merging