Re: [PATCH] fixes an issue when main thread is sleeping when receiving data.

Hmm, could this be related with the problem I was having when  
developing "membox"?

I've experienced similar freezing problems when enqueuing at high  
speed, but didn't know where the problem was. I'm definitely having a  
second look at it with this patch as well.

Regards,

Alejandro
--
Alejandro Guerrieri
[email protected]



On 30/04/2009, at 12:10, Alexander Malysh wrote:

> hi Vincent,
>
> please try attached patch. The mtbatch freeze because in main we try  
> to join all threads (gwthread_join_all) but
> in read_messages_from_bearerbox we blocked infinite waiting for acks  
> from bearerbox.
>
> Thanks,
> Alex
> <mtbatch-freeze.diff>
>
> Am 17.04.2009 um 11:35 schrieb Vincent CHAVANIS:
>
>>
>> I've experiences somes troubles when sending 3500 msg/sec
>> Sometimes, the main thread_id is freezed.
>>
>>
>> # On the code executed below:
>>
>>   sended = run_batch();
>>
>>   info(0," mark#1 ");
>>
>>   /* avoid exiting before sending all msgs */
>>   while(sended > counter_value(counter)) {
>>       info(0," mark#2 ");
>>       gwthread_sleep(0.1);
>>   }
>>
>>
>> # i got this:
>>
>> 2009-04-17 11:31:55.0720 [22720] [0] DEBUG: message length 130,  
>> sending 1 messages
>> 2009-04-17 11:31:55.0720 [22720] [0] DEBUG: message length 130,  
>> sending 1 messages
>> 2009-04-17 11:31:55.0720 [22720] [0] DEBUG: message length 130,  
>> sending 1 messages
>> 2009-04-17 11:31:55.0721 [22720] [0] DEBUG: message length 130,  
>> sending 1 messages
>> 2009-04-17 11:31:55.0721 [22720] [0] INFO: mtbatch has processed  
>> 999 messages with 0 errors
>> 2009-04-17 11:31:55.0721 [22720] [0] INFO:  mark#1
>> 2009-04-17 11:31:55.0721 [22720] [0] INFO:  mark#2
>> (freezed)
>> ^C
>>
>>
>>
>>
>> Alexander Malysh a écrit :
>>> Hi,
>>> I don't get it, why this patch is needed? Could you please  
>>> describe in detail?
>>> Thanks,
>>> Alex
>>> Am 16.04.2009 um 12:36 schrieb Vincent CHAVANIS:
>>>>
>>>> - When main thread is sleeping and receiving data by bearerbox,
>>>> we will be into an infinite loop.
>>>>
>>>> This patch fix this by waking-up the main thread.
>>>>
>>>> Vincent.
>>>>
>>>>
>>
>>
>