Skip to content

aggregator: don't push EOS on error return

This reverts b16697d2 which I don't think is justified, and creates issues for actually surfacing errors:

A regular application will often tear down a pipeline once it sees EOS on the bus, which means future buffers coming into the aggregator will not get an ERROR but a FLUSHING flow return, and the application will never know that the pipeline actually stopped because of an error.

Downstream will still have a chance to shut down cleanly upon state change.

Merge request reports