Pick an option on each card and press Done; your answer goes straight
back to the session. The raiser's pick is marked. "I'm not sure" is a real answer.
1A capped merge queue can force a session to end its turn holding a green, unarmed PR — which Fede's own standing rule forbids. What should happen instead?
In plain terms. When the merge line is full, a session can't arm its PR — Fede's rule says never leave a turn with an unarmed open PR. Should a background job take over re-arming it once there's room, should the session just wait around, or should Fede's rule bend for this one case?
Why, and what was tried
The session raising this recommends: Durable watcher re-arms it Same shape as co.propflow.mq-release, already built and running for the held-PR release: a background job owns 'wait for room, then arm' so no interactive session has to sit polling a queue count across turns. A session hits the cap, ends its turn, and the watcher arms the PR once there's room — nothing is stranded, nothing holds a turn open. Tried first, unsuccessfully: Tried: re-read #15928's own text — the doctrine already tells a capped session to 'leave the PR green and unarmed, come back later,' but names no owner for 'come back later.' Checked for an existing general-purpose re-armer: none exists outside the held-PR-specific mq-release job. Building one is a real PR, not a one-line fix, so flagging the gap rather than silently inventing the mechanism.
a4713487 is parked on this.
2A capped merge queue can force a session to end its turn holding a green, unarmed PR — which Fede's own standing rule forbids. What should happen instead?
In plain terms. When the merge line is full, a session can't arm its PR, but Fede's rule says never leave a turn with an unarmed open PR. Should a background job re-arm it later, should the session wait around, or should the rule bend here?
Why, and what was tried
The session raising this recommends: Durable watcher re-arms it Same shape as co.propflow.mq-release, already built and running for the held-PR release: a background job owns 'wait for room, then arm' so no interactive session has to sit polling a queue count across turns. Tried first, unsuccessfully: Tried: re-read #15928's own text — it tells a capped session to leave the PR unarmed and come back later, but names no owner for 'later'. No general-purpose re-armer exists outside the held-PR-specific mq-release job.
a4713487 is parked on this.
Answered and withdrawn (1)
b91001354 Coding sessions keep sending finished work to the merge line without checking how full it is, because the written rules tell them to. Should a safety check refuse to add more once the line holds 29, and should the rule change on both Gera's and Fede's machines? — answered: Gera's mini only