Ultimate LWJGL / JOGL showdown

[gouessej]

@ra4king

What compatibility problems? LWJGL doesn’t do anything except provide a direct binding to OpenGL. If you have any problems with LWJGL, blame OpenGL and the crappy drivers or blame Linux.

At first, JOGL and LWJGL are more than OpenGL bindings, they both provide their own native windowing systems.

Secondly, when there is something wrong with the OpenGL driver, it causes problems on both bindings of OpenGL and in applications using OpenGL directly in C/C++. However, Cube and Mupen 64 work fine on my both machines, they both rely on OpenGL and I’m really fed up with this kind of behaviour, it’s always the fault of OpenGL drivers! When I found a very small regression in LWJGL some years ago, the first reaction of some maintainers was to accuse the OpenGL driver whereas I succeeded in proving that it was a regression of LWJGL. OpenGL works fine on my machines, that’s all. If you believe I’m wrong, prove it.

Thirdly, “GLIBC 2.4 symbol not found” is the kind of error you get when you use too old a version of GLIBC, I have never blamed LWJGL about this. If you had some more knowledge in C, you would know that I could solve this problem by recompiling LWJGL by myself, there’s no bug.

Finally, Nicolas (N_I_C_S) can confirm I have always respected his choice of using LWJGL, that’s the same for JMonkeyEngine 2.0 despite my problems with these technologies.

@N_I_C_S

Hi Julien, thank you very much for your tests ! Yes, the slowdowns are certainly caused by the collision tests which increase exponentially with the enemies…
Otherwise, actually I’m bored of lwjgl and its compatibility problems >:( ! One of these days I’ll try to switch to jogl !

I have to be honest. LWJGL displayed the error message “GLIBC 2.4 symbol not found” whereas JOGL 2.0 crashed the whole JVM on my old machine (Celeron 700 Mhz). JOGL 1.1.1a worked fine. I succeeded in using my old graphics card on my “less old” computer.
As I already told you, JMonkeyEngine 2.0 has a better support of LWJGL, its LWJGL renderer is more stable than the JOGL renderer based on JOGL 1.1.1a despite my efforts. I know you would like to spend much time on the artificial intelligence. If you don’t want to waste a lot of time in switching between bindings and engines (like me :(), maybe go on using JMonkeyEngine 2.0 with LWJGL 2.8.2 and report any bug on the official LWJGL forum (there is no real bug tracker for this API as far as I know). Don’t make any decision in the precipitation. If you really want to give up LWJGL, of course I will spend some days to port your source code directly to Ardor3D + JOGL 2.0 so that you can stay focused on the AI.

[kappa]

@gouessej

When I found a very small regression in LWJGL some years ago, the first reaction of some maintainers was to accuse the OpenGL driver whereas I succeeded in proving that it was a regression of LWJGL. OpenGL works fine on my machines, that’s all. If you believe I’m wrong, prove it.

Ha, I remember that thread, you ranted long and hard about how LWJGL was broken and in the end it turned out that the example code you were using was broken (it didn’t set the 2d ortho mode) rather than it being an actual LWJGL bug :).

@gouessej

Thirdly, “GLIBC 2.4 symbol not found” is the kind of error you get when you use too old a version of GLIBC, I have never blamed LWJGL about this. If you had some more knowledge in C, you would know that I could solve this problem by recompiling LWJGL by myself, there’s no bug. I have to be honest. LWJGL displayed the error message “GLIBC 2.4 symbol not found” whereas JOGL 2.0 crashed the whole JVM on my old machine (Celeron 700 Mhz). JOGL 1.1.1a worked fine. I succeeded in using my old graphics card on my “less old” computer.

Any modern Linux distro should have a GLIBC version equal or better than that (After all GLIBC 2.4 was released over 10 years ago), if not, you should really be thinking about upgrading the distro rather than the recompiling LWJGL (or use an LWJGL release from that era). Also explains why JOGL1.1 works in this case.

@gouessej

report any bug on the official LWJGL forum (there is no real bug tracker for this API as far as I know).

There are not enough bug reports (nor is the project large enough) to warrant or need a full blown dedicated tracker system. The LWJGL forum has a dedicated bug tracker section which is perfectly adequate for this purpose.

[gouessej]

@kappa

Ha, I remember that thread, you ranted long and hard about how LWJGL was broken and in the end it turned out that the example code you were using was broken (it didn’t set the 2d ortho mode) rather than it being an actual LWJGL bug :).

You have just forgotten to say that this example is still now the very first demo of LWJGL on the official website. If I were you, I would not be so proud to speak about that. I spent a lot of time to understand what was wrong, please respect my efforts.

@kappa

Any modern Linux distro should have a GLIBC version equal or better than that (After all GLIBC 2.4 was released over 10 years ago), if not, you should really be thinking about upgrading the distro rather than the recompiling LWJGL (or use an LWJGL release from that era). Also explains why JOGL1.1 works in this case.

This machine is too old, it was shipped with Microsoft Windows 98, I cannot install even Mandriva Linux 2007. I use this old crap only for some tests.

@kappa

There are not enough bug reports (nor is the project large enough) to warrant or need a full blown dedicated tracker system. The LWJGL forum has a dedicated bug tracker section which is perfectly adequate for this purpose.

I’m a bit sceptic about that, lots of small projects use a bug tracker.

[kappa]

@gouessej

If I were you, I would not be so proud to speak about that. I spent a lot of time to understand what was wrong, please respect my efforts.

Fair enough, you did spend a lot of time and effort investigating the issue and the old example code was fixed, so credit is well deserved there. Only reason I mentioned it was to set the record straight because you’ve been using the same argument against the library for a while now, even though it was a relatively minor issue and not part of the actual core library.