diff options
Diffstat (limited to 'lib/tevent/tevent.h')
-rw-r--r-- | lib/tevent/tevent.h | 73 |
1 files changed, 72 insertions, 1 deletions
diff --git a/lib/tevent/tevent.h b/lib/tevent/tevent.h index 1c01a6330c..d67e2b0f20 100644 --- a/lib/tevent/tevent.h +++ b/lib/tevent/tevent.h @@ -520,9 +520,65 @@ int tevent_set_debug_stderr(struct tevent_context *ev); * requests are much easier to compose than the low-level event * handlers called from tevent_add_fd. * + * A lot of the simplicity tevent_req has brought to the notoriously + * hairy async programming came via a set of conventions that every + * async computation programmed should follow. One central piece of + * these conventions is the naming of routines and variables. + * + * Every async computation needs a name (sensibly called "computation" + * down from here). From this name quite a few naming conventions are + * derived. + * + * Every computation that requires local state needs a + * @code + * struct computation_state { + * int local_var; + * }; + * @endcode + * Even if no local variables are required, such a state struct should + * be created containing a dummy variable. Quite a few helper + * functions and macros (for example tevent_req_create()) assume such + * a state struct. + * * An async computation is started by a computation_send * function. When it is finished, its result can be received by a - * computation_recv function. + * computation_recv function. For an example how to set up an async + * computation, see the code example in the documentation for + * tevent_req_create() and tevent_req_post(). The prototypes for _send + * and _recv functions should follow some conventions: + * + * @code + * struct tevent_req *computation_send(TALLOC_CTX *mem_ctx, + * struct tevent_req *ev, + * ... further args); + * int computation_recv(struct tevent_req *req, ... further output args); + * @endcode + * + * The "int" result of computation_recv() depends on the result the + * sync version of the function would have, "int" is just an example + * here. + * + * Another important piece of the conventions is that the program flow + * is interrupted as little as possible. Because a blocking + * sub-computation requires that the flow needs to continue in a + * separate function that is the logical sequel of some computation, + * it should lexically follow sending off the blocking + * sub-computation. Setting the callback function via + * tevent_req_set_callback() requires referencing a function lexically + * below the call to tevent_req_set_callback(), forward declarations + * are required. A lot of the async computations thus begin with a + * sequence of declarations such as + * + * @code + * static void computation_step1_done(struct tevent_req *subreq); + * static void computation_step2_done(struct tevent_req *subreq); + * static void computation_step3_done(struct tevent_req *subreq); + * @endcode + * + * It really helps readability a lot to do these forward declarations, + * because the lexically sequential program flow makes the async + * computations almost as clear to read as a normal, sync program + * flow. * * It is up to the user of the async computation to talloc_free it * after it has finished. If an async computation should be aborted, @@ -587,6 +643,9 @@ typedef void (*tevent_req_fn)(struct tevent_req *req); /** * @brief Set an async request callback. * + * See the documentation of tevent_req_post() for an example how this + * is supposed to be used. + * * @param[in] req The async request to set the callback. * * @param[in] fn The callback function to set. @@ -642,6 +701,10 @@ void *tevent_req_callback_data_void(struct tevent_req *req); /** * @brief Get the private data from a tevent request structure. * + * When the tevent_req has been created by tevent_req_create, the + * result of tevent_req_data() is the state variable created by + * tevent_req_create() as a child of the req. + * * @param[in] req The structure to get the private data from. * * @param[in] type The type of the private data @@ -788,6 +851,13 @@ bool _tevent_req_cancel(struct tevent_req *req, const char *location); * req = tevent_req_create(mem_ctx, &state, struct computation_state); * @endcode * + * Tevent_req_create() creates the state variable as a talloc child of + * its result. The state variable should be used as the talloc parent + * for all temporary variables that are allocated during the async + * computation. This way, when the user of the async computation frees + * the request, the state as a talloc child will be free'd along with + * all the temporary variables hanging off the state. + * * @param[in] mem_ctx The memory context for the result. * @param[in] pstate Pointer to the private request state. * @param[in] type The name of the request. @@ -949,6 +1019,7 @@ bool _tevent_req_nomem(const void *p, * if (tevent_req_nomem(subreq, req)) { * return tevent_req_post(req, ev); * } + * tevent_req_set_callback(subreq, computation_done, req); * return req; * } * @endcode |