tcp_boundary: drop source correctly

[antiguru]

Instead of waiting to schedule ok and err streams to drop the source, only
require a single of the streams to be terminate to terminate both. This
prevents a situation where the ok stream is terminated, but the err stream
will not pick up the termination because it won't be scheduled anymore.

Signed-off-by: Moritz Hoffmann mh@materialize.com

Motivation

• This PR fixes a previously unreported bug: Sources aren't cleaned up after compute.

Testing

• This PR has adequate test coverage / QA involvement has been duly considered.

Release notes

This PR includes the following user-facing behavior changes:

• N/A

[antiguru]

This is an alternative to #11945.

[guswynn]

I am convinced now that #11945 is definitely a backwards approach, but this technique concerns me for 2 reasons:

• What is different about the ok and err stream where, as the source is dying? Is the ok stream always going to get a push? if it is also not getting any data (just like the err) stream, could neither of these be called and we have the same bug?
• My other comment inline; I don't understand timely well enough to know if this is a concern

[guswynn]

#12082 should be able to replace this pr

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[frankmcsherry]

On my list of "pending reviews" but I suspect we just want to close?

[guswynn]

@frankmcsherry yeah, this should be relevant anymore!