libs/newlib-cygwin
libs/newlib-cygwin
4
0
Fork 0
mirror of git://sourceware.org/git/newlib-cygwin.git synced 2025-01-30 02:50:25 +08:00
Code Issues Projects Releases Wiki Activity

minor cleanups

Browse Source
This commit is contained in:
Christopher Faylor 2001-09-12 04:47:47 +00:00
parent 8e4d969260
commit ff6e295ebf
2 changed files with 30 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
winsup/cygwin/ChangeLog
View File

@ -1,6 +1,7 @@
Wed Sep 12 13:03:00 2001 Robert Collins <rbtcollins@hotmail.com>
* autoload.cc (LoadDLLfuncEx): Auto load TryEnterCriticalSection - its an NT only call.
* autoload.cc (LoadDLLfuncEx): Auto load TryEnterCriticalSection - it's
an NT only call.
* thread.cc (pthread_cond::TimedWait): Use critical sections for NT.
(pthread_cond::fixup_after_fork): Don't detect bad apps.
(pthread_mutex::pthread_mutex): Use critical sections for NT.
@ -8,9 +9,11 @@ Wed Sep 12 13:03:00 2001 Robert Collins <rbtcollins@hotmail.com>
(pthread_mutex::Lock): Ditto.
(pthread_mutex::TryLock): Ditto.
(pthread_mutex::UnLock): Ditto.
(pthread_mutex::fixup_after_fork): Ditto. Also do not detect bad apps.
(__pthread_mutex_trylock): Move WIN32 specific test into the class method.
(__pthread_mutex_destroy): Prevent dereferencing passed pointer without valid address.
(pthread_mutex::fixup_after_fork): Ditto. Also do not detect bad apps.
(__pthread_mutex_trylock): Move WIN32 specific test into the class
method.
(__pthread_mutex_destroy): Prevent dereferencing passed pointer without
valid address.
* thread.h (pthread_mutex): Use critical sections for NT.
Tue Sep 11 21:55:37 2001 Christopher Faylor <cgf@cygnus.com>
@ -77,11 +80,11 @@ Sun Sep 9 20:09:11 2001 Christopher Faylor <cgf@cygnus.com>
Mon Sep 10 08:28:00 2001 Robert Collins <rbtcollins@hotmail.com>
* thread.h (MT_Interface): Remove pshared mutex array.
Add a threadsafe list for mutex tracking (for fixup after fork).
* thread.h (MT_Interface): Remove pshared mutex array. Add a
threadsafe list for mutex tracking (for fixup after fork).
* thread.cc (MTInterface::Init): Remove pshared mutex array.
(pthread_mutex::pthread_mutex): Remove pshared mutex functionality.
Fail with EINVAL on attempts to use pshared functionality.
(pthread_mutex::pthread_mutex): Remove pshared mutex functionality.
Fail with EINVAL on attempts to use pshared functionality.
(__pthread_mutex_getpshared): Remove.
(__pthread_cond_timedwait): Remove pshared mutex functionality.
(__pthread_cond_wait): Ditto.
@ -93,8 +96,8 @@ Mon Sep 10 08:28:00 2001 Robert Collins <rbtcollins@hotmail.com>
(__pthread_mutex_destroy): Ditto.
(__pthread_mutex_setprioceiling): Ditto.
(__pthread_mutexattr_setpshared): Ditto.
Sun Sep 9 23:09:00 2001 Corinna Vinschen <corinna@vinschen.de>
Sun Sep 9 23:09:00 2001 Corinna Vinschen <corinna@vinschen.de>
* pwdgrp.h (pwdgrp_check::set_last_modified): Call GetFileTime()
instead of GetFileInformationByHandle().

33
winsup/cygwin/thread.cc
View File

@ -3,6 +3,7 @@
Copyright 1998, 1999, 2000, 2001 Red Hat, Inc.
Originally written by Marco Fuykschot <marco@ddi.nl>
Substantialy enhanced by Robert Collins <<rbtcollins@hotmail.com>
This file is part of Cygwin.
@ -10,19 +11,19 @@ This software is a copyrighted work licensed under the terms of the
Cygwin license. Please consult the file "CYGWIN_LICENSE" for
details. */
/*Implementation overview and caveats:
/* Implementation overview and caveats:
Win32 puts some contraints on what can and cannot be implemented. Where possible
we work around those contrainsts. Where we cannot work around the constraints we
either pretend to be conformant, or return an error code.
Win32 puts some contraints on what can and cannot be implemented. Where
possible we work around those contrainsts. Where we cannot work around
the constraints we either pretend to be conformant, or return an error
code.
Some caveats: PROCESS_SHARED objects while they pretend to be process shared,
may not actually work. Some test cases are needed to determine win32's behaviour.
My suspicion is that the win32 handle needs to be opened with different flags for
proper operation.
Some caveats: PROCESS_SHARED objects while they pretend to be process
shared, may not actually work. Some test cases are needed to determine
win32's behaviour. My suspicion is that the win32 handle needs to be
opened with different flags for proper operation.
R.Collins, April 2001.
*/
R.Collins, April 2001. */
#ifdef HAVE_CONFIG_H
# include "config.h"
@ -449,7 +450,7 @@ pthread_cond::~pthread_cond ()
{
pthread_cond *tempcond = MT_INTERFACE->conds;
while (tempcond->next && tempcond->next != this)
tempcond = tempcond->next;
tempcond = tempcond->next;
/* but there may be a race between the loop above and this statement */
InterlockedExchangePointer (&tempcond->next, this->next);
}
@ -514,7 +515,7 @@ pthread_cond::TimedWait (DWORD dwMilliseconds)
* critical sections, which are faster, but introduce a race _here_. Until then
* The NT variant of the code is redundant.
*/
rv = SignalObjectAndWait (mutex->win32_obj_id, win32_obj_id, dwMilliseconds,
false);
#endif
@ -626,7 +627,7 @@ pthread_mutex::pthread_mutex (pthread_mutexattr *attr):verifyable_object (PTHREA
{
this->win32_obj_id =::CreateMutex (&sec_none_nih, false, NULL);
if (!win32_obj_id)
magic = 0;
magic = 0;
}
condwaits = 0;
pshared = PTHREAD_PROCESS_PRIVATE;
@ -641,7 +642,7 @@ pthread_mutex::~pthread_mutex ()
else
{
if (win32_obj_id)
CloseHandle (win32_obj_id);
CloseHandle (win32_obj_id);
win32_obj_id = NULL;
}
/* I'm not 100% sure the next bit is threadsafe. I think it is... */
@ -704,7 +705,7 @@ pthread_mutex::fixup_after_fork ()
{
win32_obj_id =::CreateMutex (&sec_none_nih, false, NULL);
if (!win32_obj_id)
api_fatal("pthread_mutex::fixup_after_fork() failed to create new win32 mutex\n");
api_fatal("pthread_mutex::fixup_after_fork() failed to create new win32 mutex\n");
}
#if DETECT_BAD_APPS
if (condwaits)
@ -746,7 +747,7 @@ semaphore::~semaphore ()
{
semaphore *tempsem = MT_INTERFACE->semaphores;
while (tempsem->next && tempsem->next != this)
tempsem = tempsem->next;
tempsem = tempsem->next;
/* but there may be a race between the loop above and this statement */
InterlockedExchangePointer (&tempsem->next, this->next);
}