Re: Re: Re: vic usability improvements

[Jin Tian <jintian@nlsde.buaa.edu.cn>]

Hi, Lloyd Lim:

  I have test these code in our project. But the conclusion is be careful to use the Tcl_DoOneEvent code in vic. I have meet error when using these code. When the source number is changing and you are titing the video window. Access Error will pop-up (The circumstance is dual CPU, that might cause the error). And I think bob's crash is same problem.

  When you are drawing a video window, and also you do OneEvent (maybe source number change). Then the engines_ chain will changed, and some of renderer pointer will be invalid. Due to no semaphore scheme is using, access error might pop-up.

  I do not know wheather there will happened on mash. But is source number increased, this thing will happened more easily.

  Maybe change the place of Tcl_DoOneEvent will be better.

======= At 2001-12-12, You wrote =======

>On Wednesday 12 December 2001 09:43 am, Jin Tian wrote:
>>
>>  I have test similiar code in vic. It seemed that with this line, vic
>> work much better under Win9x(Me), but not in Win2k. I think it is
>> because that Win2k's Message Cycle is improved. But I have not test
>> under the circumstance you metioned.
>
>My primary Windows environment is 2000, so, yes, I've tested it there 
>and the improvement is dramatic.  When 10+ Mbps are coming in, you'd 
>previously have to wait 30 seconds to a couple minutes for the user 
>interface to respond.  I'm not joking.  Now it only takes a couple 
>seconds at most.  It's not lightning fast, but it's much, much better.
>
>> Also, improving too much
>> WINDOWS_EVENT will cause Tcl Timer STARVE. So, you should confirm
>> that if you have add this line, does video rate update in tcl work at
>> every 1 second?
>
>Yes, I confirmed just now that the framerate, bitrate, and loss rate 
>numbers all update every second, just like before.
>
>>  I also test the code like GetMessage/PeekMessage loop. Vic in Win9x
>> will also working better, but I never meet any crash.
>
>I don't think Bob's crash is related to the update change.  My guess is 
>that it's related to DirectDraw.  Contrary to my announcement, 
>DirectDraw support was not enabled in the 5.1.6 release due to an 
>accidental problem with the makefile.  I enabled it last week but the 
>DirectDraw code is still being worked on.
>
>Lloyd

= = = = = = = = = = = = = = = = = = = =

  Best Wishes to u.

==================================================
	Jin Tian(½ðÌì)
	PhD. Student
	Advance Network Technologies Group (Admire Group)
	http://www.nlsde.buaa.edu.cn/projects/admire/
	National Lab of Software Development Environment
	Beijing University of Aeronautics and Astronautics
	P.B.: 2-83# 37 Xueyuan Road  Beijing 100083  P.R.China
	E.B.: jintian@nlsde.buaa.edu.cn
	P.N.: (86)(10)82317603(O)
==================================================