16 Custodians

When an extension allocates resources that must be explicitly freed (in the same way that a port must be explicitly closed), a Racket object associated with the resource should be placed into the management of the current custodian with scheme_add_managed.

Before allocating the resource, call scheme_custodian_check_available to ensure that the relevant custodian is not already shut down. If it is, scheme_custodian_check_available will raise an exception. If the custodian is shut down when scheme_add_managed is called, the close function provided to scheme_add_managed will be called immediately, and no exception will be reported.

Scheme_Custodian*

 

scheme_make_custodian

(

Scheme_Custodian* m)

Creates a new custodian as a subordinate of m. If m is NULL, then the main custodian is used as the new custodian’s supervisor. Do not use NULL for m unless you intend to create an especially privileged custodian.

Scheme_Custodian_Reference*

scheme_add_managed

(

Scheme_Custodian* m,

 

 

Scheme_Object* o,

 

 

Scheme_Close_Custodian_Client* f,

 

 

void* data,

 

 

int strong)

Places the value o into the management of the custodian m. If m is NULL, the current custodian is used.

The f function is called by the custodian if it is ever asked to “shutdown” its values; o and data are passed on to f, which has the type

typedef void (*Scheme_Close_Custodian_Client)(Scheme_Object *o,

                                              void *data);

If strong is non-zero, then the newly managed value will be remembered until either the custodian shuts it down or scheme_remove_managed is called. If strong is zero, the value is allowed to be garbage collected (and automatically removed from the custodian).

Independent of whether strong is zero, the value o is initially weakly held. A value associated with a custodian can therefore be finalized via will executors.

The return value from scheme_add_managed can be used to refer to the value’s custodian later in a call to scheme_remove_managed. A value can be registered with at most one custodian.

If m (or the current custodian if m is NULL)is shut down, then f is called immediately, and the result is NULL.

void

 

scheme_custodian_check_available

(

Scheme_Custodian* m,

 

 

 

 

const char* name,

 

 

 

 

const char* resname)

Checks whether m is already shut down, and raises an error if so. If m is NULL, the current custodian is used. The name argument is used for error reporting. The resname argument will likely be used for checking pre-set limits in the future; pre-set limits will have symbolic names, and the resname string will be compared to the symbols.

void

 

scheme_remove_managed

(

Scheme_Custodian_Reference* mref,

 

 

 

 

Scheme_Object* o)

Removes o from the management of its custodian. The mref argument must be a value returned by scheme_add_managed or NULL.

void

 

scheme_close_managed

(

Scheme_Custodian* m)

Instructs the custodian m to shutdown all of its managed values.

void

 

scheme_add_atexit_closer

(

Scheme_Exit_Closer_Func f)

Installs a function to be called on each custodian-registered item and its closer when Racket is about to exit. The registered function has the type

  typedef

  void (*Scheme_Exit_Closer_Func)(Scheme_Object *o,

                                  Scheme_Close_Custodian_Client *f,

                                  void *d);

where d is the second argument for f.

At-exit functions are run in reverse of the order that they are added. An at-exit function is initially registered (and therefore runs last) that flushes each file-stream output port.

An at-exit function should not necessarily apply the closer function for every object that it is given. In particular, shutting down a file-stream output port would disable the flushing action of the final at-exit function. Typically, an at-exit function ignores most objects while handling a specific type of object that requires a specific clean-up action before the OS-level process terminates.