This feature is obsolete. Although it is still supported by browsers, its usage is discouraged in new projects. Try to avoid using it.
JSObjectOps is not a supported API. Details of the API may change from one release to the next. This documentation should be considered SpiderMonkey internals documentation, not API documentation. See bug 408416 for details.
typedef void (*JSPropertyRefOp)(JSContext *cx, JSObject *obj, JSProperty *prop);
||A context that was the
||The object of which
||The property to release. That is, the value that
The following contract governs
JSObjectOps callers and implementations:
JSPropertypointer, the property is locked.
- While the property is locked, the caller may access it using other
JSObjectOpscallbacks. It may not otherwise call into the JSAPI in ways that might trigger other property accesses. (In a
JS_THREADSAFEbuild, that would risk deadlock.)
- When finished with a locked property, the caller must release it by calling the
- The lifetime of the
JSProperty *extends from the successful
lookupPropertycall to the
As a SpiderMonkey implementation detail, what is actually locked here (under the native implementation of
JSObjectOps) is not the property but some collection of objects including the object on which the property is defined. The granularity of the locking is up to the
JSObjectOps implementation; deadlock does not happen because each thread accesses only one property at a time.
A single, built-in
JSObjectOps implementation is used for most SpiderMonkey objects that are exposed to scripts. Custom
JSObjectOps implementations can either retain the SpiderMonkey property storage and locking scheme (by copying all or most of the built-in
JSObjectOps) or replace it entirely.
Threads. In a
JS_THREADSAFE build, any consistency observed by multiple threads operating on the same data is provided solely by the property locking scheme described above. In SpiderMonkey 1.8 and earlier, the main built-in
JSObjectOps implementation serialized all accesses to a given object's properties. (That is, for each object, all property accesses happened in some order, and what each thread observed was consistent with that order: no stale reads, for example.) However, SpiderMonkey does not guarantee this high degree of serialization.