See Custodians for basic information on the Racket custodian model.
In GRacket, eventspaces managed by cust are also shut down.
Closes all file-stream ports, TCP ports, TCP listeners, and UDP sockets that are managed by cust (and its subordinates), and empties all custodian boxes associated with cust (and its subordinates). It also removes cust (and its subordinates) as managers of all threads; when a thread has no managers, it is killed (or suspended; see thread/suspend-to-kill) If the current thread is to be killed, all other shut-down actions take place before killing the thread.
In GRacket, custodians also manage eventspaces.
Memory accounting is normally available in Racket 3m, which is the main variant of Racket, and not normally available in Racket CGC.
Returns #t if Racket is compiled with support for per-custodian memory accounting, #f otherwise.
(custodian-require-memory limit-cust need-amt stop-cust) → void? limit-cust : custodian? need-amt : exact-nonnegative-integer? stop-cust : custodian?
If a check is registered, and if Racket later reaches a state after garbage collection (see Garbage Collection) where allocating need-amt bytes charged to limit-cust would fail or trigger some shutdown, then stop-cust is shut down.
(custodian-limit-memory limit-cust limit-amt [ stop-cust]) → void? limit-cust : custodian? limit-amt : exact-nonnegative-integer? stop-cust : custodian? = limit-cust
If a check is registered, and if Racket later reaches a state after garbage collection (see Garbage Collection) where limit-cust owns more than limit-amt bytes, then stop-cust is shut down.
A custodian’s limit is checked only after a garbage collection, except that it may also be checked during certain large allocations that are individually larger than the custodian’s limit. A single garbage collection may shut down multiple custodians, even if shutting down only one of the custodians would have reduced memory use for other custodians.
For reliable shutdown, limit-amt for custodian-limit-memory must be much lower than the total amount of memory available (minus the size of memory that is potentially used and not charged to limit-cust). Moreover, if individual allocations that are initially charged to limit-cust can be arbitrarily large, then stop-cust must be the same as limit-cust, so that excessively large immediate allocations can be rejected with an exn:fail:out-of-memory exception.