Patch #483: get_message_from_server in a thread

[Vegar]

What it does is creating a thread that listens for messages from the server and puts them in a queue which is emptied in the main loop. I wrote it because all the calls to get_message_from_server() in the drawing functions are really ugly.

 

It has been on berlios for a couple weeks now.

Has anyone tested it? Did it work well? Any comments/suggestions?

[Zep]

Has anyone tested it? Did it work well? Any comments/suggestions?

177293[/snapback]

 

Ive been using it for a week now, On various connections.

Its very nice on high latency connections,

And i no longer see the rendering loop stalled by data related issues on slower systems.

 

It should defininatly be added to mainline IMHO

[Vegar]

Ive been using it for a week now, On various connections.

Its very nice on high latency connections,

And i no longer see the rendering loop stalled by data related issues on slower systems.

177298[/snapback]

I'm glad to hear it's working the way it should be

[Learner]

I'm glad to hear it's working the way it should be 

177300[/snapback]

Since the packets are added to the queue and not emptied until the main loop is hit, it may not help systems avoid resyncs if they video board is a bottleneck. But, it will help the high latency connections and possibly during major dance parties as well.

[Placid]

Perhaps a choice in EL.ini in that case?

[Wytter]

There won't be any disadvantages with it really... So all it needs is testing.

[Vegar]

Commited it to CVS. Compile with -DNETWORK_THREAD to enable it.

[Cajs]

Seems to do a pretty good work of avoiding resynchs for me (avg ping time is 350-450ms for me right now since my isp is having problems so I was getting plenty of them...)

 

On the other hand there seems to be a problem with EL running in the background (e.g. with browser window with msg board over it) - if you are in a busy area like a storage EL crashes ussually within 15-20 seconds of me openning another window over it.

 

This is from non-debug of elc.exe as I can't seem to get the debug one compiling ever since cal3d was put in...

 

Error:

Unhandled exception at 0x00409a3c in elc.exe: 0xC0000005: Access violation reading location 0x09ba3000.

 

It is happening while freeing a block in the heap with

void free(pblock) - free a block in the heap

 

will try to get more info once my debug version of client starts compiling again...

~C

[Lachesis]

Couldn't reproduce this Anyone else?