[yswdqz]

Fixes #4498

Changes

Align BatchLogRecordProcessor batching behavior with the
BatchSpanProcessor fix in #4466.

• Export at most one batch per normal worker iteration instead of
  draining the queue in a tight loop.
• Wait for a full batch unless the scheduled delay expires or an
  explicit wakeup occurs. Existing queue-pressure wakeups are preserved.
• Snapshot the buffered record count for ForceFlush and shutdown,
  exporting records in batches capped by max_export_batch_size.
• Avoid treating subsequent exports as force-flush operations after
  a previous ForceFlush has completed.
• Add regression tests for full-batch waiting, half-full queue wakeups,
  flush/shutdown batch limits, and behavior after ForceFlush.

No public API changes.

Testing

Validated locally on macOS with AppleClang, C++17, and a Debug build:

• All 18 BatchLogRecordProcessor tests passed.
• clang-format 18 checks passed for the modified C++ files.
• markdownlint-cli 0.46.0 passed for CHANGELOG.md.

The full CI matrix has not been run locally.

Checklist

• CHANGELOG.md updated for non-trivial changes
• Unit tests have been added
• Changes in public API reviewed — no public API changes

[owent]

LGTM, thanks

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.51%. Comparing base (3ee38aa) to head (f3f3dbb).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4743      +/-   ##
==========================================
- Coverage   86.51%   86.51%   -0.00%     
==========================================
  Files         530      530              
  Lines       21138    21136       -2     
==========================================
- Hits        18286    18284       -2     
  Misses       2852     2852              

Files with missing lines	Coverage Δ
sdk/src/logs/batch_log_record_processor.cc	91.01% <100.00%> (-0.09%)	⬇️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[mateenali66]

checked whether this needs the locked notify from #4741, since my note there said logs would once the full-batch wait landed. it doesn't. OnEmit sets is_force_wakeup_background_worker before it notifies, so a half-full notify ends the wait instead of re-parking on a partial batch the way the span worker does.

stress probe with queue 2, batch 2 and a 10 min delay, one emit, a swept spin, then two more emits, 50k rounds per run. the worker sleeps through a full queue in 15 and 19 rounds on main and in 15 and 17 at d23cb0c, so that stall is older than this PR and unchanged by it. the #4741 change mirrored into OnEmit takes it to 0 of 50k. that fits a follow-up once both merge, not a blocker here.