sleep-poll

A deny rule by default: it refuses a sleep before a status read, a call that only sleeps, or a wait on or retry into a full budget, and names what to run instead.

What it catches

A sleep before a status read, a call that only sleeps, or a wait on or retry into a full budget.

Why

Both are a wait whose loop is the caller's own next call. A sleep of ten seconds or more followed on the same line by a status read (kill -0, the process table, magus status or a job row, a pull request's checks, a CI run) looks once and leaves the next look to the next call. A call whose only work is sleeping, printing and reading a file once (sleep 300; echo tick, sleep 120 && tail -5 out.log, a for loop of sleeps and echoes) is a timer: its notice says only that time passed, where the work's own backgrounded call is announced with its result. Backgrounded it is refused at any length; in the foreground from ten seconds, since a foreground sleep past the tool timeout is moved to the background and becomes one. A shorter foreground sleep is a settle, and a line that also builds, tests or runs a script is work. A watch (--watch) is not a status read, and a loop is busy-wait's. A worker (a session whose lease is a child row or a subagent's) waits for the build budget only in the call the wait gates: magus status --wait piped into a magus call, which reserves the seat for it (a run, an affected or a graph build), or followed by && and the call. A magus status --wait that gates no magus call, --watch, a status read beside a sleep or inside a watcher, or a magus run after a sleep of ten seconds or more is refused. A worker that cannot wait commits, reports to its orchestrator that its check found no seat, and stops; the orchestrator waits once for every worker, and a spawn or magus job fork into a full budget tells it how full the budget is and which run has held it longest. A magus run or magus affected whose argv matches one this checkout's run log recorded refused with MGS3009 in the last 60 seconds is the same poll with no sleep at all: it is refused with magus status --wait | <the command> as the next line, which waits for a seat and then runs, or for a worker with the report-and-stop above. A line whose status --wait gates the run passes, and only run logs written inside the minute are opened. It compares argv exactly, so the same run with a flag added, removed or reordered is a different command. It cannot tell a one-off settle before a single check from the tenth look of a poll; the served watch answers both.

Default and override

By default this rule takes the decision deny. A workspace sets it by name, in its root magusfile, to deny, advise or off:

magus\guard.builtins({"sleep-poll": "advise"})

A loosening takes effect once it is committed; a tightening applies at once.

Seeing it

A verdict names its rule in brackets, which is how you got here:

deny [sleep-poll]: ...

A workspace that sets this rule to advise sees advise [sleep-poll]: ... instead.

magus describe rule sleep-poll prints the same entry at a terminal, and magus describe rules lists every rule this workspace enforces.

See also

• All rules - what this workspace enforces, deny first
• The guard - how a verdict is reached and wired
• Setting a built-in rule - how a workspace changes this default