Heya, Many thanks for your swift reply. I have been through my code and cleaned up the few instances without proper masking when calling beginEditCP etc.
Current status, however, remains : 1) Standalone application : Initialisation works fine 2) 1x Server, 1x Client : Initialisations works from both client and distributed, and server self-initialising. All subsequent initialising from the client *always* works. 3) 2x Server 1x Client : Initialisation only occurs within the first server, nothing occurs on the second server, but this is inconsistent. Sometimes it doesn't initialise at all. 4) 3x Server 1x Client : Initialisation occurs on the second server (!), the first and third server do no initialise. 5) 4x Server 1x Client : Initialisation occurs on the second server and on none of the other servers. In all of the above cases, as I have said, if I initialise on the client I have no problems because the scenegraph is treated as normal. However, if I attempt to simply to distribute the core parameters as part of the node core, and then initialise at the server side to reduce transfer of tons of geometry, it is inconsistent. I am using "cout <<" statements to confirm initialisation, as well as looking for the geometry on screen. One concern is that perhaps the geometry is being clipped/optimised or such before it reaches renderEnter for that core, is this possible? It is, as you correctly guessed, derived from Geometry, I have uploaded 3x images to show how it looks when it works correctly, and when it fails on 2 and 4 servers http://www.flashygraphics.co.uk/images/heightMap.png http://www.flashygraphics.co.uk/images/2servers.png http://www.flashygraphics.co.uk/images/4servers.png Any ideas? Craig Moore Research Officer University of Brighton EPOCH Dirk Reiners <[EMAIL PROTECTED]> wrote: > > Hi Craig, > > [EMAIL PROTECTED] wrote: > > Heya, > > > > Long time reader, first time writer to the lists. I am currently trying > > to get to grips with the development of custom nodes, and am currently > > working on my second test node. The main purpose of the node is to : > > > ... > > > > Here is where my question comes in. I have added a quick test to the > > renderEnter function that tests whether the core has been initialised or > > not by calling > > > > if (!getInitialised()) { > > initialiseFromImage(); > > } > > > > This means that if for whatever reason, after I edited some details and > > decided to invalidate the mesh, and decided to mark it as uninitialised, > > it would regenerate the code when the renderAction activates the > > renderEnter function. > > > > This has a slightly unexpected effect. When I set up a single server, > > and a single client, and run the client through all of the steps above > > and I *don't* call geo->initialiseFromImage();, the server receives the > > image, mesh resolution and initialisation data, but not the mesh data, > > and when it tries to render on the server, it notices that the data > > hasn't been initialised, and calls initialiseFromImage() from the > > server, so I don't need to distribute very large amounts of geometry. I > > have tested and this seems to work, however, only with a single server. > > If I attempt to use two different servers (and these are the generic > > servers, simply compiled from the OpenSG tutorial), neither of them > > register that my core has not been initialised, the client does not > > initialise, and so in fact no geometry is ever created. It is as though > > the core has not been distributed with the correct value of "initialised". > > > > Is there something fundamentally missing in my understanding about > > distribution works, I am missing something? Because with initialisation > > happening on the client, it works perfectly, and running it without > > distribution works perfectly. > > Nope, that sounds perfectly fine. There is really no reason for one or two > servers to behave differently. > > One scenario that I could imagine causing the behavior: I assume your NodeCore > > is derived from Geometry. If you call beginEdit on it without setting a mask, > the inherited parts will be transfered, too, which would overwrite the changes > > made on the server and remove the created geometry. Can you try setting the > fieldmasks for begin/endEdit and see if that makes a difference? > > Dirk > > ------------------------------------------------------------------------- > This SF.net email is sponsored by DB2 Express > Download DB2 Express C - the FREE version of DB2 express and take > control of your XML. No limits. Just data. Click to get it now. > http://sourceforge.net/powerbar/db2/ > _______________________________________________ > Opensg-users mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/opensg-users > ------------------------------------------------------------------------- This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ _______________________________________________ Opensg-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/opensg-users
