[sentry-junior]

When log_flush_threshold is not set, logs are kept in a buffer of 1000. Once it's full, each new log pushes out the oldest one, which is never sent. This records each of those as log_item / buffer_overflow.

The buffer doesn't say when it drops an item, so the change compares the count before and after the push. A full buffer keeps the same count. An unbounded buffer (log_flush_threshold set) always grows, so it never records anything. The test fills the buffer, checks that nothing was recorded yet, then adds one more log.

I used buffer_overflow because the client reports spec defines it for SDK-internal buffers, and the span recorder limit (#2309) already uses it. The Python SDK records its log batcher overflow as queue_overflow instead.

Part of #2228.

via Martin Linzmayer.

--

View Session [Sentry]