Main Page Namespace List Class Hierarchy Alphabetical List Compound List File List Namespace Members Compound Members File Members Related Pages
ACE_Process_Manager Class Reference
Manages a group of processes.
More...
#include <Process_Manager.h>
Inheritance diagram for ACE_Process_Manager:
[legend]Collaboration diagram for ACE_Process_Manager:
[legend]List of all members.
|
Public Types |
enum | { DEFAULT_SIZE = 100
} |
Public Methods |
| ACE_Process_Manager (size_t size=ACE_Process_Manager::DEFAULT_SIZE, ACE_Reactor *reactor=0) |
int | open (size_t size=DEFAULT_SIZE, ACE_Reactor *r=0) |
int | close (void) |
| Release all resources. Do not wait for processes to exit.
|
virtual | ~ACE_Process_Manager (void) |
pid_t | spawn (ACE_Process *proc, ACE_Process_Options &options) |
pid_t | spawn (ACE_Process_Options &options) |
int | spawn_n (size_t n, ACE_Process_Options &options, pid_t *child_pids=0) |
int | wait (const ACE_Time_Value &timeout=ACE_Time_Value::max_time) |
pid_t | wait (pid_t pid, const ACE_Time_Value &timeout, ACE_exitcode *status=0) |
pid_t | wait (pid_t pid, ACE_exitcode *status=0) |
int | reap (pid_t pid=-1, ACE_exitcode *stat_loc=0, int options=WNOHANG) |
int | register_handler (ACE_Event_Handler *event_handler, pid_t pid=ACE_INVALID_PID) |
int | remove (pid_t pid) |
int | terminate (pid_t pid) |
int | terminate (pid_t pid, int sig) |
size_t | managed (void) const |
| Return the number of managed Processes.
|
void | dump (void) const |
| Dump the state of an object.
|
Static Public Methods |
ACE_Process_Manager * | instance (void) |
| Get pointer to a process-wide <ACE_Process_Manager>.
|
ACE_Process_Manager * | instance (ACE_Process_Manager *) |
void | close_singleton (void) |
| Delete the dynamically allocated singleton.
|
void | cleanup (void *instance, void *arg) |
Public Attributes |
| ACE_ALLOC_HOOK_DECLARE |
| Declare the dynamic allocation hooks.
|
Protected Methods |
virtual int | handle_signal (int signum, siginfo_t *=0, ucontext_t *=0) |
Private Methods |
int | resize (size_t) |
| Resize the pool of Process_Descriptors.
|
ssize_t | find_proc (pid_t process_id) |
ssize_t | find_proc (ACE_HANDLE process_handle) |
int | insert_proc (ACE_Process *process) |
int | append_proc (ACE_Process *process) |
int | remove_proc (size_t n) |
int | notify_proc_handler (size_t n, ACE_exitcode status) |
Private Attributes |
ACE_Process_Descriptor * | process_table_ |
| Vector that describes process state within the Process_Manager.
|
size_t | max_process_table_size_ |
size_t | current_count_ |
| Current number of processes we are managing.
|
ACE_Event_Handler * | default_exit_handler_ |
ACE_Recursive_Thread_Mutex | lock_ |
| This lock protects access/ops on <process_table_>.
|
Static Private Attributes |
ACE_Process_Manager * | instance_ = 0 |
| Singleton pointer.
|
int | delete_instance_ = 0 |
Friends |
class | ACE_Process_Control |
Detailed Description
Manages a group of processes.
This class allows applications to control groups of processes, similar to how the <ACE_Thread_Manager> controls groups of threads. Naturally, it doesn't work at all on platforms, such as VxWorks or pSoS, that don't support process. There are two (main) ways of using <ACE_Process_Manager>, depending on how involved you wish to be with the termination of managed <ACE_Process>es. If you just want <Process>es to go away when they're finished, simply register the <Process_Manager> with an <ACE_Reactor>: ACE_Process_Manager mgr( 100, some_reactor ) -or- ACE_Process_Manager mgr; ... mgr.open( 100, some_reactor ); Then, the <Process_Manager> will clean up after any <Process>es that it spawns. (On Unix, this means executing a wait(2) to collect the exit status -- and avoid zombie processes; on Win32, it means closing the process and thread HANDLEs that are created when CreateProcess is called.) If, on the other hand (and for some inexplicable reason) you want to explicitly invoke the terminated <Process> cleanup code, then *don't* register the <Process_Manager> with a Reactor, and be sure to call one of the <Process_Manager::wait> functions whenever there might be managed <Process>es that have exited. Note that in either case, <Process_Manager> allows you to register "<Event_Handlers>" to be called when a specific <Process> exits, or when any <Process> without a specific <Event_Handler> exits. When a <Process> exits, the appropriate <Event_Handler>'s <handle_input> is called; the <ACE_HANDLE> passed is either the Process' HANDLE (on Win32), or its pid cast to an <ACE_HANDLE> (on unix). It is also possible to call the <Process_Manager::wait> functions even though the <Process_Manager> is registered with a <Reactor>. Note also that the wait functions are "sloppy" on Unix, because there's no good way to wait for a subset of the children of a process. The wait functions may end up collecting the exit status of a process that's not managed by the <Process_Manager> whose <wait> you invoked. It's best to only use a single <Process_Manager>, and to create all subprocesses by calling that <Process_Manager>'s <spawn> method. Incidentally, when you register your <Process_Manager> with a <Reactor> its notification pipe is used to help "reap" the available exit statuses. Therefore, you must not use a <Reactor> whose notify pipe has been disabled. Here's the sequence of steps used to reap the exit statuses in this case: + The <Process_Manager> registers a signal handler for SIGCHLD. + The SIGCHLD handler, when invoked, uses the <Reactor>'s <notify> method to inform the <Reactor> to wake up. + Next, the <Reactor> calls the <Process_Manager>'s <handle_input>, this happens synchronously, not in sighandler-space. + The <handle_input> method collects all available exit statuses.
Member Enumeration Documentation
Constructor & Destructor Documentation
ACE_Process_Manager::ACE_Process_Manager |
( |
size_t |
size = ACE_Process_Manager::DEFAULT_SIZE, |
|
|
ACE_Reactor * |
reactor = 0 |
|
) |
|
|
|
Initialize an <ACE_Process_Manager> with a table containing up to <size> processes. This table resizes itself automatically as needed. If a non-NULL <reactor> is provided, this <ACE_Process_Manager> uses it to notify an application when a process it controls exits. By default, however, we don't use an <ACE_Reactor>. |
ACE_Process_Manager::~ACE_Process_Manager |
( |
void |
|
) |
[virtual] |
|
|
Destructor releases all resources and does not wait for processes to exit. |
Member Function Documentation
int ACE_Process_Manager::append_proc |
( |
ACE_Process * |
process |
) |
[private] |
|
|
Append information about a process, i.e., its <process_id> in the <process_table_>. Each entry is added at the end, growing the table if necessary. |
void ACE_Process_Manager::cleanup |
( |
void * |
instance, |
|
|
void * |
arg |
|
) |
[static] |
|
int ACE_Process_Manager::close |
( |
void |
|
) |
|
|
|
Release all resources. Do not wait for processes to exit.
|
void ACE_Process_Manager::close_singleton |
( |
void |
|
) |
[static] |
|
|
Delete the dynamically allocated singleton.
|
void ACE_Process_Manager::dump |
( |
void |
|
) |
const |
|
|
Dump the state of an object.
|
ssize_t ACE_Process_Manager::find_proc |
( |
ACE_HANDLE |
process_handle |
) |
[private] |
|
|
Locate the index of the table slot occupied by <process_handle>. Returns ~0 if <process_handle> is not in the <process_table_> |
ssize_t ACE_Process_Manager::find_proc |
( |
pid_t |
process_id |
) |
[private] |
|
|
Locate the index of the table slot occupied by <process_id>. Returns -1 if <process_id> is not in the <process_table_> |
int ACE_Process_Manager::handle_signal |
( |
int |
signum, |
|
|
siginfo_t * |
= 0, |
|
|
ucontext_t * |
= 0 |
|
) |
[protected, virtual] |
|
|
On Unix, this routine is called asynchronously when a SIGCHLD is received. We just tweak the reactor so that it'll call back our <handle_input> function, which allows us to handle Process exits synchronously.
On Win32, this routine is called synchronously, and is passed the HANDLE of the Process that exited, so we can do all our work here
Reimplemented from ACE_Event_Handler. |
int ACE_Process_Manager::insert_proc |
( |
ACE_Process * |
process |
) |
[private] |
|
|
Insert a process in the table (checks for duplicates). Omitting the process handle won't work on Win32... |
ACE_Process_Manager * ACE_Process_Manager::instance |
( |
ACE_Process_Manager * |
|
) |
[static] |
|
|
Set pointer to a process-wide <ACE_Process_Manager> and return existing pointer. |
ACE_Process_Manager * ACE_Process_Manager::instance |
( |
void |
|
) |
[static] |
|
|
Get pointer to a process-wide <ACE_Process_Manager>.
|
ACE_INLINE size_t ACE_Process_Manager::managed |
( |
void |
|
) |
const |
|
|
Return the number of managed Processes.
|
int ACE_Process_Manager::notify_proc_handler |
( |
size_t |
n, |
|
|
ACE_exitcode |
status |
|
) |
[private] |
|
|
If there's a specific handler for the Process at index <n> in the table, or there's a default handler, call it. |
int ACE_Process_Manager::open |
( |
size_t |
size = DEFAULT_SIZE, |
|
|
ACE_Reactor * |
r = 0 |
|
) |
|
|
|
Initialize an <ACE_Process_Manager> with a table containing up to <size> processes. This table resizes itself automatically as needed. If a non-NULL <reactor> is provided, this <ACE_Process_Manager> uses it to notify an application when a process it controls exits. By default, however, we don't use an <ACE_Reactor>. |
int ACE_Process_Manager::reap |
( |
pid_t |
pid = -1, |
|
|
ACE_exitcode * |
stat_loc = 0, |
|
|
int |
options = WNOHANG |
|
) |
|
|
|
Reap the result of a single process by calling <ACE_OS::waitpid>, therefore, this method is not portable to Win32. If the child is successfully reaped, <remove> is called automatically. This method does the same thing that the <wait> method directly above it does -- It's just here for backwards compatibility. |
|
Register an Event_Handler to be called back when the specified process exits. If pid == ACE_INVALID_PID this handler is called when any process with no specific handler exits. |
int ACE_Process_Manager::remove |
( |
pid_t |
pid |
) |
|
|
|
Remove process <pid> from the table. This is called automatically by the <reap> method after it successfully reaped a <SIGCHLD> signal. It's also possible to call this method directly from a signal handler, but don't call both <reap> and <remove>! |
int ACE_Process_Manager::remove_proc |
( |
size_t |
n |
) |
[private] |
|
|
Actually removes the process at index <n> from the table. This method must be called with locks held. |
int ACE_Process_Manager::resize |
( |
size_t |
|
) |
[private] |
|
|
Resize the pool of Process_Descriptors.
|
|
Create a new process by passing <options> to <ACE_Process::spawn>. On success, returns the process id of the child that was created. On failure, returns ACE_INVALID_PID. |
|
Create a new process by passing <options> to <proc.spawn>. On success, returns the process id of the child that was created. On failure, returns ACE_INVALID_PID. |
|
Create <n> new processes by passing <options> to <ACE_Process::spawn>, which is called <n> times. If <child_pids> is non-0 it is expected to be an array of <n> <pid_t>'s, which are filled in with the process ids of each newly created process. Returns 0 on success and -1 on failure. |
int ACE_Process_Manager::terminate |
( |
pid_t |
pid, |
|
|
int |
sig |
|
) |
|
|
|
On OSs that support signals, send the signal to the specified process. Returns 0 on success and -1 on failure. |
int ACE_Process_Manager::terminate |
( |
pid_t |
pid |
) |
|
|
|
Abruptly terminate a single process with id <pid> using the <ACE::terminate_process> method. Note that this call is potentially dangerous to use since the process being terminated may not have a chance to cleanup before it shuts down. Returns 0 on success and -1 on failure. |
|
Wait indefinitely for a single process to terminate. If pid==0, waits for any of the managed <Process>es (but see the note in the class documentation for caveats about this -- "sloppy Process cleanup on unix") If pid != 0, waits for that <Process> only. Returns the pid of the process whose exit was handled, or ACE_INVALID_PID on error. |
|
Wait up to <timeout> for a single process to terminate. If pid==0, waits for any of the managed <Process>es (but see the note in the class documentation above for caveats about this -- "sloppy process cleanup on unix") If pid != 0, waits for that <Process> only. Returns the pid of the Process whose exit was handled, 0 if a timeout occurred, or ACE_INVALID_PID on error. |
|
Block until there are no more child processes running that were <spawn>ed by this <ACE_Process_Manager>. Unlike the <wait> call below, this method does not require a signal handler or <ACE_OS::sigwait> because it simply blocks synchronously waiting for all the children managed by this <ACE_Process_Manager> to exit. Note that this does not return any status information about the success or failure of exiting child processes, although any registered exit_handlers are called. Returns 0 on success (and <remove>s the corresponding <ACE_Process_Descriptor> entries from the <Process_Manager>; otherwise, returns -1 on failure. |
Friends And Related Function Documentation
friend class ACE_Process_Control [friend]
|
|
Member Data Documentation
ACE_Process_Manager::ACE_ALLOC_HOOK_DECLARE
|
|
|
Declare the dynamic allocation hooks.
|
size_t ACE_Process_Manager::current_count_ [private]
|
|
|
Current number of processes we are managing.
|
|
This event handler is used to notify when a process we control exits. |
int ACE_Process_Manager::delete_instance_ = 0 [static, private]
|
|
|
Controls whether the <Process_Manager> is deleted when we shut down (we can only delete it safely if we created it!) |
ACE_Process_Manager * ACE_Process_Manager::instance_ = 0 [static, private]
|
|
|
This lock protects access/ops on <process_table_>.
|
size_t ACE_Process_Manager::max_process_table_size_ [private]
|
|
|
Maximum number of processes we can manage (should be dynamically allocated). |
|
Vector that describes process state within the Process_Manager.
|
The documentation for this class was generated from the following files:
Generated on Wed Jan 14 22:46:45 2004 for ACE by
1.2.18