You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 0dc711e
Browse filesBrowse the repository at this point in the historyBrowse files
Stage 2b review round 6: close the publication window, and stop trusting retirement
Checking whether the terminal is still ours and publishing the prepared frame
were two steps. A writer abandoning the reservation between them had its
retirement overwritten by the publication that followed, leaving a frame in
flight that nothing had invalidated. They are one transaction now.
Commit asks about the terminal's state again rather than inferring it from
retirement. Removing that guard in the first review round was wrong: a
publication racing a retirement produces a frame that matches on identity and
on every generation, so identity carries the argument only while nothing else
can publish. It can.
0 commit comments