Magnus Svensson wrote:
> tor 2007-10-04 klockan 11:11 +0200 skrev Mats Kindahl:
>
>> Hi Magnus!
>>
>> Patch approved. I checked that the recursion count really is increased.
>>
>
> Good, I did that in the debugger. Couldn't really figure out how to
> implement that assert you mentioned to be atomic. Or should I do it only
> only as a test?
>
Not really critical to have an assertion, as long as we have verified
that it is a documented behavior. Microsoft isn't likely to change it.
> The scary thing is that we really are looking into a struct defined by
> MS.
>
>
>> Just my few cents,
>> Mats Kindahl
>>
>> msvensson@stripped wrote:
>>
>>> Below is the list of changes that have just been committed into a local
>>> 5.0 repository of msvensson. When msvensson does a push these changes will
>>> be propagated to the main repository and, within 24 hours after the
>>> push, to the public repository.
>>> For information on how to access the public repository
>>> see http://dev.mysql.com/doc/mysql/en/installing-source-tree.html
>>>
>>> ChangeSet@stripped, 2007-10-03 21:38:32+02:00, msvensson@stripped +2 -0
>>> Bug#30992 Wrong implementation of pthread_mutex_trylock()
>>> It's not possible to use WaitForSingleObject to wait
>>> on a CRITICAL_SECTION, instead use the TryEnterCriticalSection function.
>>> - if "mutex" was already taken => return EBUSY
>>> - if "mutex" was aquired => return 0
>>>
>>> include/config-win.h@stripped, 2007-10-03 21:38:30+02:00,
> msvensson@stripped +3 -0
>>> Make windows.h define TryEnterCriticalSection
>>>
>>> mysys/my_winthread.c@stripped, 2007-10-03 21:38:30+02:00,
> msvensson@stripped +10 -12
>>> Use the TryEnterCriticalSection function to implement
> pthread_mutex_trylock
>>> Prevent recursive behaviour by looking at the RecursionCount variable
>>> in the CRITICAL_SECTION struct and return EBUSY from function if
>>> the mutex was already locked once.
>>>
>>> diff -Nrup a/include/config-win.h b/include/config-win.h
>>> --- a/include/config-win.h 2007-03-28 23:00:25 +02:00
>>> +++ b/include/config-win.h 2007-10-03 21:38:30 +02:00
>>> @@ -19,6 +19,9 @@
>>> /* We have to do this define before including windows.h to get the AWE API
>>> functions */
>>> #define _WIN32_WINNT 0x0500
>>> +#else
>>> +/* Get NT 4.0 functions */
>>> +#define _WIN32_WINNT 0x0400
>>> #endif
>>>
>>> #if defined(_MSC_VER) && _MSC_VER >= 1400
>>> diff -Nrup a/mysys/my_winthread.c b/mysys/my_winthread.c
>>> --- a/mysys/my_winthread.c 2007-10-01 15:11:12 +02:00
>>> +++ b/mysys/my_winthread.c 2007-10-03 21:38:30 +02:00
>>> @@ -40,30 +40,28 @@ void win_pthread_init(void)
>>> pthread_mutex_init(&THR_LOCK_thread,MY_MUTEX_INIT_FAST);
>>> }
>>>
>>> +
>>> /**
>>> Adapter to @c pthread_mutex_trylock()
>>>
>>> @retval 0 Mutex was acquired
>>> @retval EBUSY Mutex was already locked by a thread
>>> - @retval EINVAL Mutex could not be acquired due to other error
>>> */
>>> int
>>> win_pthread_mutex_trylock(pthread_mutex_t *mutex)
>>> {
>>> - switch (WaitForSingleObject(mutex, 0)) {
>>> - case WAIT_TIMEOUT:
>>> - return EBUSY;
>>> -
>>> - default:
>>> - case WAIT_FAILURE:
>>> - return EINVAL;
>>> -
>>> - case WAIT_OBJECT_0:
>>> - case WAIT_ABANDONED: /* The mutex was acquired because it was
>>> - * abandoned */
>>> + if (TryEnterCriticalSection(mutex))
>>> + {
>>> + /* Don't allow recursive lock */
>>> + if (mutex->RecursionCount > 1){
>>> + LeaveCriticalSection(mutex);
>>> + return EBUSY;
>>> + }
>>> return 0;
>>> }
>>> + return EBUSY;
>>> }
>>> +
>>>
>>> /*
>>> ** We have tried to use '_beginthreadex' instead of '_beginthread' here
>>>
>>>
>>>
>>
--
Mats Kindahl
Lead Software Developer
Replication Team
MySQL AB, www.mysql.com
