Linux graphics driver

Hi,

I'm currently testing a JOGL application on several Linux distributions. One of them does not support the proprietary graphics driver and is running the standard Linux driver. The application works OK in the main part. However, while trying to find the source of one issue, I noticed that the buffers that have been created using glGenBuffers are not being deleted when glDeleteBuffers is used.

Is the standard Linux graphics driver generally not suitable for JOGL?  If not, is there anything that can be used when proprietary drivers are not available?

Thanks.

Hi

Sven told me recently that glDeleteBuffers works as expected but maybe I should check it by myself. If it occurs only with a specific driver, it doesn't come from JOGL. Some open source drivers work reliably, don't conclude that this bug is present in all open source drivers.

Thanks. I'll see if I can report it somewhere.

Can you tell me which driver is concerned by this bug please? I feel concerned too as my game creates a lot of VBOs and I use an open source Xorg driver on my main computer.

No problem.  I've reported the bug here:

https://bugzilla.redhat.com/show_bug.cgi?id=968445

That details the driver etc.

I also tried this with a freeglut program (located as an attachment in the bug report).  This had the same problem so the issue was certainly not with JOGL.

Thanks.

Thank you very much. Is this bug still reproducible when you use the name at least once, when calling glBindBuffer?

Yes it is.  It was the application I was testing that led me to the issue in the first place.  This obviously performs drawing etc. using the buffers.  I just tried to make the test method as simple as possible.

Hi,

I continuing to test an application on different Linux distributions. I'm having major problems with the new distributions where the proprietary graphics driver for my graphics card is not supported. However, I was hoping that I'd be able to get somewhere with the open source driver.

On Fedora 18 x86_64, Kubuntu 13.04 x86_64 (so far), I can't even get a simple JOGL application to display at all. This image shows the output of a simple JOGL application that just draws a couple of shapes.



I think this may be a driver issue. However, I can run a simple freeglut program on these systems successfully and that made me wonder. Is this a driver issue?

Thanks,

Hi

My game has been tested under Ubuntu and Lubuntu with success. I use Mageia Linux at home and I tested with Fedora, it worked like a charm. Please run the unit tests provided with JOGL and enable all logs. In my humble opinion, it is a bug reproducible with a very few open source drivers under a very few Linux distros, there is no need to use a scary title and please be more accurate, indicate the exact driver version number and the graphics card in use instead of vaguely claiming you use an open source driver (proprietary drivers have bugs too). I have used JOGL since 2006 mainly under GNU Linux often with open source drivers (proprietary ones were not available any more in most cases) and I confirm it works like a charm. This time your problem may come from JOGL but we need some information. Please write a very complete bug report with your test case, all logs and the details about the machine you use.

Edit.: I've merged your latest thread with the previous one because they are both about problems with a very few open source drivers under GNU Linux and I hate scaring inaccurate titles.

Hi,

I think that's a bit over the top to be honest.  My first post was entitled "Linux graphics driver", hardly scary.  It was deliberately generic, because it was meant to be a generic question. I have little experience with the open source drivers and didn't want to post bug reports that were not real. I also didn't want to spend lots of time trying to locate bugs, if the open source drivers were known to be buggy or not good with jOGL, hence asking the question. As you stated they are generally good, I posted a bug report with Red Hat and also did some work in freeglut to further prove the case.

My second post could have a title that indicates the driver in use. However, I still don't think it's scary. I also believe I'm  not making any claims, just stating facts. The text inside clearly states the distributions that have been trialed. Also, I'm not sure how you can conclude this affects "very few Linux distros". I have tried 2 at random and both give similar results on my machine. Neither of us know for sure, but I think that shows there are potentially others that would be the same.

As you believe the latest issue may concern JOGL, I'll perform the recommendations you have made and post a bug report as requested.

Apologies if my posts have raised unnecessary concern. They were not meant to be bug reports, but requests for help and opinion. I was actually trying not to waste peoples time.

Thanks.

Hi,

Is there somewhere that details the unit tests you would like to see results from?

Thanks,

Please follow these instructions:
http://jogamp.org/wiki/index.php/Jogl_FAQ#Bugreports_.26_Testing

You can find most of our tests here:
https://github.com/sgothel/jogl/tree/master/src/test/com/jogamp/opengl/test

Maybe you should report bugs about the driver upstream, directly against Xorg.

Good news on the GLDeleteBuffers. Xorg have confirmed that this is the intended behaviour.  Rather than search for the lowest buffer name, when GLGenBuffers is called, it returns the next available name.  This is implemented in this manner for speed.

I have posted a bug report for the second problem I listed.

Thanks for the help.

gsxruk wrote

Good news on the GLDeleteBuffers. Xorg have confirmed that this is the intended behaviour.  Rather than search for the lowest buffer name, when GLGenBuffers is called, it returns the next available name.  This is implemented in this manner for speed.

Thank you for the feedback. As soon as "released" buffer identifiers become really usable later, it is ok for me. Do you have a reference about that piece of information?

gsxruk wrote

I have posted a bug report for the second problem I listed.

Thanks for the help.

Ok, thanks. I hope that you have been very accurate because I have never had such troubles with any open source drivers and I use them very often with ATI, Nvidia and Intel graphics chips.

Xorg say that if the names reached the maximum, further calls should start to search for available buffers from the past used buffers.  However, the number is so large, they doubt it happens very often, if at all.

I've tried to be as accurate as possible.  I am concerned it's another false bug, but I can't get it to work in any of the distributions I've tried.  But the freeglut program does work on all of them.  If you let me know if there's any other information required, I'm happy to help where I can.

Thanks.