Known issues

Currently there is a single known issue: a design change was fairly recently made, in the process of formalizing the spec, where it was decided that a 'happy abort' (ie. a cancellation under expected circumstances, like the successful completion of a process that did not need to consume all upstream data) should be exposed by the pipeline as an EndOfStream (as if the source stream ran out of data entirely) instead of an Aborted.

You may run into some older implementations of source streams that still implement the old behaviour, as this change has not been applied everywhere yet. Please report it when this happens, and they will be fixed promptly!