functionality with RTLEvent:
o the thread manager semaphore functionality isn't used for anything else,
and the main difference between mutex/condvar and semaphores is that the
latter can be used for IPC (while the thread manager is always used
within a single process) and that they are counting (but since a thread
can only suspend itself nowadays, it cannot be suspended recursively)
o OS X sandboxing does not support the use of semaphores (or rather: Apple
does not allow sandboxing rules that enable the use of IPC semaphores in
the appstore)
git-svn-id: trunk@28965 -
Especially for the introduction of the Queue method the internal Synchronize handling was modified. Instead of handling only one event there is now a queue of events which is walked completely when CheckSynchronize is called. Each entry in the queue can carry a PRTLEvent which will be signaled when the contained method has been executed and thus Synchronize methods can still be blocking.
Exceptions inside the queued methods are either handed back to the calling method for Synchronize events or raised directly (after leaving the queue in a valid state) to the caller of CheckSynchronize.
The way platform specific adjustments can be made to TThread was changed. Instead of implementing the Constructor and Destructor directly one now implements the methods SysCreate and SysDestroy which are called from the Constructor and Destructor respectively. All RTLs were adjusted for this and should be controlled by the platform maintainers for correct compilation (Unix works).
The new method NameThreadForDebugging has two overloaded variants: one with the thread name as AnsiString and one with the thread name as UnicodeString.
By default the AnsiString variant calls the UnicodeString variant and the latter needs to be implemented. This can be changed by defining THREADNAME_IS_ANSISTRING for a platform. Then the UnicodeString variant calls the AnsiString one and the AnsiString one needs to be implemented.
Also added was a global property CPUCount for the System unit. This property returns the number of virtual cores of the system.
New methods and functions that should be implemented per platform are:
System.GetCPUCount (default returns 1)
Classes.TThread.GetSystemTimes (default zeros the struct)
Classes.TThread.NameThreadForDebugging (default does nothing)
More detailed information about the added methods will be available in the feature announcement mail.
git-svn-id: trunk@23227 -
race conditions in case DoTerminate can cause the thread to be freed
already. Side-effect: changing FreeOnTerminate in DoTerminate no longer
has any effects, but this is compatible with the generic code for
ThreadProc in objpas/classes/classes.inc (patch by Nikolai ZHUBR)
git-svn-id: trunk@14504 -
for qwordbool) + test:
o assigning true to such a variable now sets them to $ff/$ffff/$ffffffff
o these types are now all signed
o converting an integer type to a byte/word/long/qwordbool using an
explicit type cast keeps the integer's original value stored in the
bool, instead of forcing it to ord(true)/ord(false)
(mantis #10233 and #10613, implemented for all architectures, testsuite
tested for ppc32, sparc and x86)
* fixed some places where the rtl depended on longbool(true) having the
value 1
* extended several boolean tests (and adapted some to no longer assume
that byte/word/long/qwordbool(true)=1)
+ support for converting to qwordbool in second_int_to_bool for x86, ppc
and sparc
git-svn-id: trunk@9898 -
(Vinzent Hoefler)
* simplified CWaitForThreadTerminate based on comments from Vinzent
Hoefler
* fixed resource leaks where in some cases a pthread would not be
reaped based on comments from Vinzent Hoefler (resolves#9016)
git-svn-id: trunk@7588 -
on unix
* fixed some dataraces in unix' suspend/resume (although they can't
be solved completely due to no possibility to atomically
suspend/resume and set fsuspended at the same time)
git-svn-id: trunk@5677 -
sem_open/sem_close for Darwin (doesn't have sem_init/sem_destroy)
+ implementation of cSemaphore* based on pipes (for potential future
systems that don't have either sem* routines)
+ test for basicrtlevent
* fixed datarace whereby a TThread could be started, run and exit before
TThread.AfterConstructor had been called (Mantis 6693, all platforms)
* throw EThread exceptions in TThread.create if something during creating
the tthread goes wrong (*nix)
* don't crash in TThread.Destroy if the TThread throws an exception before
it was fully initialised (*nix)
* changed order of operations in TThread.Destroy so it doesn't perform
invalid thread operations in some edge cases (*nix)
* fixed usage of sem_wait/sem_trywait (can be interrupted) in Semaphore
and RTLEvent implementations
* fixed erroneous waiting for threads after they had already exited via
pthread_detach/pthread_exit
* fixed several memory leaks in case of thread intialisation errors
(*nix)
* unified tthread.inc for all Unices
git-svn-id: trunk@5662 -
