ofFBO crashing on allocate

[jvcleave]

MacPro1,1 ATI Radeon X1900 XT:

Using this code

void testApp::setup(){
    ofSetLogLevel(OF_LOG_VERBOSE);
    fboObject.allocate(500, 500);
}

Console reads OF: OF_LOG_NOTICE: ofFbo::checkGLSupport() maxColorAttachments: 4 maxDrawBuffers: 4 maxSamples: 6

I get a EXC_BAD_ACCESS error on line 370 of ofFBO.cpp

Here is a screenshot of the debugger http://imbx.us/Xld.png

another with ofFBO::Settings expanded http://imbx.us/X9f1.png

[ofTheo]

hmm - quick check. does it work if the fbo is 512 x 512 ?

[jvcleave]

I am on a different machine (same model MacPro1,1 but with a NVIDIA GeForce 7300 GT) but it is still crashing in the same place (width/height set to 500 or 512).

OF: OF_LOG_NOTICE: ofFbo::checkGLSupport() maxColorAttachments: 4 maxDrawBuffers: 4 maxSamples: 0

I can re-check on the original machine in a few hours

[ofTheo]

can you try this:

//--------------------------------------------------------------
void testApp::setup(){
    fbo.allocate(500, 500, GL_RGBA);
}

//--------------------------------------------------------------
void testApp::update(){

}

//--------------------------------------------------------------
void testApp::draw(){

    fbo.begin();
        ofClear(255, 255, 255, 255);
        ofSetColor(255, 0, 0);
        ofCircle(mouseX, mouseY, 100);
    fbo.end();

    ofSetColor(255);
    fbo.draw(0, 0);
}

[ofTheo]

ps - did you pull from the latest git? also make sure you have ofFbo fbo; in testApp.h

[jvcleave]

I pulled earlier today and just ran the above code and got the same crash.

I just pulled again

From https://github.com/openframeworks/openFrameworks 263718d..ef806b1 master -> origin/master Updating 263718d..ef806b1

and cleaned all targets, ran again and got the same error

[jvcleave]

I've uploaded the test project here in case it helps https://github.com/jvcleave/FBOTest

[ofTheo]

hmm what OS are you running on ? can you post detailed info about your setup?

also maybe try ofDisableArbTex() at the beginning of testApp setup and try sizes of 512 512

[jvcleave]

Still crashing :/

Is there a 0062 project you would like me to try?

OS X 10.6.8

Model Name: Mac Pro Model Identifier: MacPro1,1 Processor Name: Dual-Core Intel Xeon Processor Speed: 2.66 GHz Number Of Processors: 2 Total Number Of Cores: 4 L2 Cache (per processor): 4 MB Memory: 7 GB Bus Speed: 1.33 GHz Boot ROM Version: MP11.005C.B08 SMC Version (system): 1.7f10

NVIDIA GeForce 7300 GT:

Chipset Model: NVIDIA GeForce 7300 GT Type: GPU Bus: PCIe Slot: Slot-1 PCIe Lane Width: x16 VRAM (Total): 256 MB Vendor: NVIDIA (0x10de) Device ID: 0x0393 Revision ID: 0x00a1 ROM Revision: 3008 Displays: Cinema: Resolution: 1680 x 1050 Pixel Depth: 32-Bit Color (ARGB8888) Display Serial Number: 2A7022K2UFZ Main Display: Yes Mirror: Off Online: Yes Rotation: Supported BenQ FP202W: Resolution: 1680 x 1050 @ 60 Hz Pixel Depth: 32-Bit Color (ARGB8888) Mirror: Off Online: Yes Rotation: Supported

[jvcleave]

I uploaded a 0062 version that uses ofxFbo that works with your example

https://github.com/jvcleave/FBOTest_0065

[ofTheo]

@memo @arturoc can you take a look at this?

[arturoc]

is working for me without problem, which line are you getting the crash? the line 370 which you tell about in the first comment is a comment in the current version

[arturoc]

i had some modifications in my code, sorry, so line 370 is:

glGenFramebuffers(1, &fbo);

right? is super weird that is crashing there as there's no heap memory allocation or nothing similar there. can you upload somewhere a screenshot of the backtrace to see some more info?

[jvcleave]

yeah - that is the line. I just tested on my Macbook Pro 2011 without any errors so may it's something with older cards?

Here are screens of me stepping through - is that what you needed or is there a way of getting more info? (sorry - finding a host for the images)

[jvcleave]

Here is a set of the screenshots http://www.flickr.com/photos/jvcleave/5953145814/in/set-72157627228766992/lightbox/

[jvcleave]

Just did this.

in testApp.h

GLuint fbo;

and in testApp::setup

glGenFramebuffers(1, &fbo);

007 crashed, 0062 was fine

[jvcleave]

So it looks like my version of 007 is using glGenFramebuffers from openFrameworks/libs/glew/include/GL/glew.h

and my 0062 install is using glGenFramebuffers from /System/Library/Frameworks/OpenGL.framework/Versions/A/Headers/glext.h

[ofTheo]

oh interesting. in testApp::setup can you try putting this line first?

glewInit();

[arturoc]

mmh, glewInit should have been called by then. does fbo's work at all in that computer with 0062?

i think that function is null because it's not supported, probably we need to check if it exists before trying to call it

[arturoc]

can you test with the last commit? i've added some functionality to check if the graphics card supports fbos

[ofTheo]

but for him it does exist - in 0062 its being defined by: /System/Library/Frameworks/OpenGL.framework/Versions/A/Headers/glext.h

and that works.

it could be a bug with glew setting a different address than glext

[ofTheo]

hmm could it be related to this: http://www.gamedev.net/topic/570462-glgenframebuffers-null-pointer-in-32-context-but-works-with-31/page__p__4644630#entry4644630

can you also try this: http://www.gamedev.net/topic/570462-glgenframebuffers-null-pointer-in-32-context-but-works-with-31/page__p__4644653#entry4644653

[arturoc]

yes, i guess so, just wanted to test if gl is detecting the extension at all.

also can you see what's the address of that function? i guess you can just see it by passing the mouse over it or adding a watch in the debugger

El mar, 19-07-2011 a las 05:29 -0700, ofTheo escribió:

but for him it does exist - in 0062 its being defined by: /System/Library/Frameworks/OpenGL.framework/Versions/A/Headers/glext.h

and that works.

it could be a bug with glew setting a different address than glext

[arturoc]

yeah, that seems like it, try adding:

glewExperimental= GL_TRUE;

before glewInit in ofAppRunner.cpp line 96

[jvcleave]

Test 1 NVIDIA GeForce 7300 GT

pulled OF

From https://github.com/openframeworks/openFrameworks ef806b1..51ffdb7 master -> origin/master Updating ef806b1..51ffdb7

Cleaned all targets

Still crashes

Console output:

OF: OF_VERBOSE: FBO supported OF: OF_LOG_NOTICE: ofFbo::checkGLSupport() maxColorAttachments: 4 maxDrawBuffers: 4 maxSamples: 0

glew.h leads to openFrameworks/libs/glew/include/GL/glew.h via Apple+clicking glGenFramebuffers in Xcode

//=====================

Test 2 NVIDIA GeForce 7300 GT

void testApp::setup(){ ofSetLogLevel(OF_LOG_VERBOSE); glewInit(); fbo.allocate(500, 500, GL_RGBA); }

Still crashes

I then re-ran with a breakpoint on glewInit and get this while stepping in the debugger

1 0x0001f82e in _glewInit_GL_VERSION_1_2 at iostream:77

//===================== Test 3

Changed openFrameworks/libs/openFrameworks/app/ofAppRunner.cpp

void ofSetupOpenGL(ofPtr windowPtr, int w, int h, int screenMode){ window = windowPtr; window->setupOpenGL(w, h, screenMode); glewExperimental= GL_TRUE;

ifndef TARGET_OPENGLES

GLenum err = glewInit();
if (GLEW_OK != err)
{
    /\* Problem: glewInit failed, something is seriously wrong. */
    ofLog(OF_LOG_ERROR, "Error: %s\n", glewGetErrorString(err));
}

endif

ofSetCurrentRenderer(ofPtr<ofBaseRenderer>(new ofGLRenderer(false)));
//Default colors etc are now in ofGraphics - ofSetupGraphicDefaults
//ofSetupGraphicDefaults();

}

Cleaned all targets

SUCCESS! TEST RAN!

Console output OF: OF_VERBOSE: FBO supported OF: OF_LOG_NOTICE: ofFbo::checkGLSupport() maxColorAttachments: 4 maxDrawBuffers: 4 maxSamples: 0 OF: OF_LOG_NOTICE: FRAMEBUFFER_COMPLETE - OK

Thanks guys! will test on other machine as well - need anything else?

[arturoc]

cool, thanks, have just uploaded the fix

[jvcleave]

Just wanted to confirm this also worked on the original system with the ATI Radeon X1900 XT.

Thanks again!