Mouse cursor doesnt show correctly(updating from 0.5 to 0.6)

[eugen]

Hi! This is my first time on this forum, CEGUI is great
One thing doesnt work after updating from 0.5 to 0.6 and that is the fact that the mouse cursor doesnt look correctly. I tried it with different scheme files (which i got from the CEGUI site, i suppose the default ones: Taharez, Valnilla) and it seems it just doesnt show the correct part of the overall texture image.
What could be the problem?!
(the other controls seem not to show correctly in all cases but i can clearly see windows and buttons, push buttons and see the other state images)

[CrazyEddie]

Hi, and welcome

Not sure what the issue could be at the moment; IIRC nothing affecting that kind of thing changed much, anyway, a few questions...

1a) Which renderer / engine is this using?
1b) If OpenGL, which Image codec module?
2a) Which operating system?
2b) If you are on Windows, do you use our SDKs or did you build CEGUI yourself?
3) Did you (or can you) build the sample apps? Do they show correctly?

Thanks for the extra info, it will hopefully help to narrow down the potential causes

CE.

[eugen]

Hi, Thanks for assisting.
1a. im using ogre, dx
2a. Windows
2b. ive built CEGUI myself, upgrading from 0.5 to 0.6 by diffing i know here can be a big problem but is strange that is working correctly everything else but the images
3. i havent built the sample apps

[CrazyEddie]

Hmmm. This is very odd. The issue is that I can't explain why upgrading would break anything, especially with Ogre, since the renderer is in the Ogre tree and so is the exact same code.

The only thing I can do at the moment is tell you, ignoring the upgrade from 0.5.0 to 0.6.0, what would normally come to mind in this situation.

Normally issues with imagesets and imagery not being in the 'right' place according to what is defined in the XML are caused by non power of two textures getting stretched when loaded and throwing out the pixel positions. Though that makes no sense if the same thing is happening with images that are power of two (i.e. our imagesets for Taharez and Vanilla).

Can you get the texture for the imageset file as loaded by Ogre and see what has actually been loaded - is it correctly loaded, up the right way, etc...

As you can probably tell, I'm kind of pulling ideas out of the air here

CE.

[eugen]

Yes, im sorry i have so little data to give you about my problem, i will be cheking out what you suggested and get back. Im sort of tied up to our development right now and i dont have enough time to check this as deep as i should,
I realy apreciate your help and i thank you very much for taking your time to do this!

this is an image showing the error...the white small sqare is the cursor. The buttons i think they are not quite alright also



this is one with taharez look, it seems this one is clearly screwed on more than the mouse cursor. In this image the cursor is on the right just above the window title



these are the only cegui related messages i get into the ogre log

"Texture: _cegui_ogre_0: Loading 1 faces(PF_A8R8G8B8,256x256x1) with 0 generated mipmaps from Image. Internal format is PF_A8R8G8B8,256x256x1.
"
and
"Texture: TaharezLook.tga: Loading 1 faces(PF_A8R8G8B8,256x256x1) with hardware generated mipmaps from Image. Internal format is PF_A8R8G8B8,256x256x1.
"

All scheme and config files are from samples/datafiles from CEGUI 0.6 release.

Judging from the results i have with all the schemes it seems is not related to config, image or scheme files. Is smth global and maybe could have smth to do with the code?

[CrazyEddie]

Hi,

Thanks for the screen shots; it helps to clarify things immensely.

Obviously all the imagery from the loaded Imagesets is all over the place, though the fact that text is showing correctly indicates that the Imageset systems do function correctly since the font/text system is built on top of it (albeit with dynamically created imagesets), this means that the issue is almost certainly related to image loading.

Having looked at what is showing instead of what should be showing and comparing to the source images, nothing in particular is indicated.

Do you get the same result from the OpenGL driver under Ogre?

Anyhow, I will try to test this later today (though it may end up being done at the weekend) on Windows - to see if I can come up with any ideas. Is it 1.4.8 Ogre? And presumably D3D 9?

CE

[eugen]

Is the same with OpenGL in Ogre also. Yes, i was using dx9 for the shots.
The Ogre version is 1.4.7 but i suppose there arent noticeable changes to 1.4.8 in the way the images are loaded
Thanks!

[CrazyEddie]

Hi,

Thanks for the extra info.

It's certainly odd. I think if it was a common issue somebody else would have mentioned it by now.

Anyway, I could not try anything out yesterday but will possibly try today

CE.

[CrazyEddie]

I did some basic testing on Windows using Ogre 1.4.7 and v0-6 branch code (basically what will become 0.6.1, though no relevant changes have been made from 0.6.0 that would affect any tests).

I integrated the new CEGUI into Ogre and ran the Ogre sample apps that use cegui, with the Ogre supplied CEGUI datafiles - no issues there.

I then integrated Ogre and the updated Ogre CEGUI renderer into the CEGUI samples and ran those, using the CEGUI supplied datafiles - no issues there.

Is it just like your 'main app' that you get the issue with, if you have such a thing, or does a minimal example also demonstrate the issues?

Do you use a different scene manager or anything of that nature?

Do you use a stock Ogre, or are there mods?

Since you upgraded CEGUI in what might be considered an 'unusual' way (using diffs), does this mean you had existing mods to CEGUI? If so what were/are they? In either case, how does the CEGUI you're using differ now from the 0.6.0 release code?

There's obviously something going on, while I think it's definitely not "us", there is an outside possibility of some strangeness or possible missing render setting in the CEGUI renderer module for Ogre (hence, some of the questions). Failing that it must be something modified or app/client side (hence the rest of the questions!).

If you're convinced it's not something local to your own app (and I guess you would be if it only broke due to the upgrade), what would be most useful right now is a minimal example that demonstrates the issue

CE.

[eugen]

First of all thanks for all the involment.
The 'problem' with examples is that we have here a pretty complex project structure with alot of other modules and all are tied together so it is not easy to make a 'simple' example to show the problem.
We are not using a different scene manager, only octree.
We use modified ogre but like CEGUI code modifications are just by inheriting some interfaces. There is no relevant Ogre or CEGUI code changed. Of course, somewhere should be a problem since we cant get it right but normally our changes shouldnt affect Ogre or CEGUI default behavior.
Im pretty sure now that is smth wrong with our versions here and the code. I will go on again with the diff to see if there is smth ive forgot.
What can i tell you now is when switching versions i had a few minor issues moving to the new CEGUI renderer and xml parser. We use expat now, falagard and default image plugin from CEGUI.

I will go once again with the diff. The changes to CEGUI is only in the base, no other modules were changed.

[CrazyEddie]

First of all thanks for all the involment.

No problemo It's certainly a puzzler! Currently I am coming up empty as far as to a cause.

Please keep us informed, I'd like to know the eventual outcome / cause. And of course if a solution does come to me in the mean time, I'll reply here

CE.

[Jon]

Hmm, I wonder if I am seeing something similar. I'm using the same code base for both Ogre and CEGUI.

Edit: um nevermind. There was a width of 10 in both imagesets, but it wasn't as apparent under the Demo.

[eugen]

This is strange. I found the time to actually replace the current modified version i had here with a clean 0.6 version after whitch i added the changes i did for the former version. (like i said, the changes are only some classes inheriting some general porpose interrfaces). The result is the same after recompiling the whole engine. Which is pretty dissapointing for me cause i had actually expected it to be fixed

so the problem should lie into the rendering engine and its integration with the CEGUI. Not that is a problem with Ogre but with my version on Ogre and the new CEGUI.
[edit]
i kinda take it for granted that i dont need to update the OgreGUIRenderer to have it working with 0.6 version. I suppose im correct since all compile fine

if u have any suggestions of testing this more deeply ill be more than welcome to try them. Im not sure where to start testing this, where to look in more detail. Also if u have some simple test cases which will show relevant information about one thing or another i will test them...

And if u can think of smth to test I think we should go more in the engine direction (meaning no test cases on CEGUI alone with the examples and all, since they are not realy relevant for what i have here)

For example i could take a look in debug when the image is mapped on a simple window so i can see what is wrong with the coordinates (if this is possible and mostly where is this code) or some other simple tests we can do to reveal more about the problem.

What do u think?

[CrazyEddie]

i kinda take it for granted that i dont need to update the OgreGUIRenderer to have it working with 0.6 version. I suppose im correct since all compile fine

This is an incorrect assumption. You absolutely must rebuild the Ogre CEGUI renderer module against the new CEGUI libs - anything else is a definite recipe for disaster

For example i could take a look in debug when the image is mapped on a simple window so i can see what is wrong with the coordinates (if this is possible and mostly where is this code) or some other simple tests we can do to reveal more about the problem.

Yes, I think this kind of test would be ideal, and is absolutely the only place to start in trying to trace the issue (IMO) - what you want to do is get Ogre to either draw the Ogre::Texture that is underlying the CEGUI::Imageset to screen, or better yet write it to a file, then you'll be able to see exactly what the imagery is looking like internally.

I'm not sure how to do that in Ogre without looking it up (so you can look it up instead!) - if all else fails I guess you could use CEGUI to render the entire texture to the screen by defining an Imageset / Image that covers the entire texture area and outputting it to the screen (though you did say you wanted an non-CEGUI test.).

Definitely let us know how this goes, please

CE.

[eugen]

This is an incorrect assumption. You absolutely must rebuild the Ogre CEGUI renderer module against the new CEGUI libs - anything else is a definite recipe for disaster
CE.

i meant i didnt need to get any new code for the CEGUI renderer, of course i built it against the new CEGUI version

i will have a look more deeply in the next days and ill keep posting here the results.

Thanks for your support!