Sundar Yamunachari wrote:
> Sarah Jelinek wrote:
>>
>>
>> Sundar Yamunachari wrote:
>>> Sarah Jelinek wrote:
>>>> Hi Sundar and Jan,
>>>>
>>>> Please review the changes for:
>>>>
>>>> 533 Caiman crashes if transfer fails during install (out of memory?)
>>>>
>>>> webrev is located at:
>>>>
>>>> http://cr.opensolaris.org/~sjelinek/bug_533/
>>>>
>>>> Regards,
>>>> sarah
>>>>
>>>> _______________________________________________
>>>> caiman-discuss mailing list
>>>> caiman-discuss at opensolaris.org
>>>> http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
>>>>   
>>> perform_slim_install.c:
>>>
>>> 97: Remove the line. "ti_done" is no longer used.
>> will do.
> Also  lines 616 and 624.
>>>
>>> 451: If you are going to wait for ti to complete, why create a 
>>> thread? Why not directly call libti?
>> Because we still want to run the TI processing in its own thread. We 
>> have progress reporting we need to have run in its own thread. The 
>> pthread_join doesn't stop the thread from running, and it still 
>> enables to gui to reap the progress reporting. It doesn't block this 
>> mechanism where if we called libti directly we would block the 
>> progress reporting. pthread_join simply allows me to get the exit 
>> status of the do_ti function so I can properly handle an error.
> Are we sending multiple callbacks to GUI for TI? I see Orchestrator is 
> sending one callback at the end of TI processing. Is TI module sending 
> callbacks directly to GUI? If we plan to send multiple callbacks to 
> GUI to report TI progress in the future, then it makes sense to create 
> a thread.
We send multiple callbacks. Look at ti_cb(). We actually report for 3 
milestones.

sarah
****
>
> - Sundar
>>
>> The other thing I could do is call libti directly and put the 
>> progress reporting for TI in a separate thread. This way seemed the 
>> most straightforward way to fix this issue. But, I am open to 
>> discussion on this.
>>
>> thanks for the review,
>> sarah
>> ****
>>
>>
>>>
>>> - Sundar
>>>
>
>

Reply via email to