|
|
|
| AllegroGL and Intel chipsets. |
|
Richard Phipps
Member #1,632
November 2001
 |
great. thanks! EDIT: Here is the AllegroGL error log when trying to open a screen with a via/s3g deltachrome chipset: Quote:
al-main INFO: Allegro initialised (instance 1)
The pixel format is invalid. (0xc00707d0)
Format 1: (0x00000006) Format 2: (0x00000006) Format 3: (0x00000006) Format 4: (0x00000006) Format 5: (0x00000006) Format 6: (0x00000006) Format 7: (0x00000006) Format 8: (0x00000006) Format 9: (0x00000006) Format 10: (0x00000006) Format 11: (0x00000006) Format 12: (0x00000006) Format 13: (0x00000006) Format 14: (0x00000006) Format 15: (0x00000006) Format 16: (0x00000006) Format 17: (0x00000006) Format 18: (0x00000006) Format 19: (0x00000006) Format 20: (0x00000006) Format 21: (0x00000006) Format 22: (0x00000006) Format 23: (0x00000006) Format 24: (0x00000006) Format 25: (0x00000006) Format 26: (0x00000006) Format 27: (0x00000006) Format 28: (0x00000006) Format 29: (0x00000006) Format 30: (0x00000006) Format 31: (0x00000006) Format 32: (0x00000006) Format 33: (0x00000006) Format 34: (0x00000006) Format 35: (0x00000006) Format 36: (0x00000006) Format 37: (0x00000006) Format 38: (0x00000006) Format 39: (0x00000006) Format 40: (0x00000006) Format 41: (0x00000006) Format 42: (0x00000006) Format 43: (0x00000006) Format 44: (0x00000006) Format 45: (0x00000006) Format 46: (0x00000006) Format 47: (0x00000006) Format 48: (0x00000006) Format 49: (0x00000006) Format 50: (0x00000006) Format 51: (0x00000006) Format 52: (0x00000006) Format 53: (0x00000006) Format 54: (0x00000006) Format 55: (0x00000006) Format 56: (0x00000006) Format 57: (0x00000006) Format 58: (0x00000006) Format 59: (0x00000006) Format 60: (0x00000006) Format 61: (0x00000006) Format 62: (0x00000006) Format 63: (0x00000006) Format 64: (0x00000006) Format 65: (0x00000006) Format 66: (0x00000006) Format 67: (0x00000006) Format 68: (0x00000006) Format 69: (0x00000006) Format 70: (0x00000006) Format 71: (0x00000006) Format 72: (0x00000006) Format 73: (0x00000006) Format 74: (0x00000006) Format 75: (0x00000006) Format 76: (0x00000006) Format 77: (0x00000006) Format 78: (0x00000006) Format 79: (0x00000006) Format 80: (0x00000006) Format 81: (0x00000006) Format 82: (0x00000006) Format 83: (0x00000006) Format 84: (0x00000006)
al-gfx ERROR: Failed setting graphic driver 1330072663.
So, any ideas..? |
|
Milan Mimica
Member #3,877
September 2003
 |
Use the SVN version. The is a known problem with some Intel chipsets, at least I know about it. We had a discussion but not a real fix.
-- |
|
Richard Phipps
Member #1,632
November 2001
|
I'll look into that, seems that log is from the old AllegroGL version. I downloaded the SVN version yesterday and we found that it crashed on Intel 915 and 940 chipsets as before. Disabling the ARB_multitexturing extension worked though. (I don't use multitexturing in the game) I thought the latest SVN was supposed to fix that Intel bug? |
|
Bob
Free Market Evangelist
September 2000
 |
Quote: Format 1:
The handle is invalid. This looks like a real AGL bug. Quote: Disabling the ARB_multitexturing extension worked though. Don't you mean ARB_multisampling? They're not quite the same... -- |
|
Richard Phipps
Member #1,632
November 2001
|
Sorry, multisampling. It's the same issue as before: Quote:
if (allegro_gl_extensions_GL.ARB_multisample) { Apparently, the variable allegro_gl_extensions_GL suggests that the Intel 915series chipset is capable of handling this Open GL extension, but it is not. I suspect that the extensions are being parsed improperly, but I ran out of time to verify this today. I tried the SVN version (yesterday) and the crashing was still there. So I disabled the relevant multisample parts. We've also just confirmed that the same issue affects the via/s3g deltachrome chipset. Commenting out multisampling makes it work, with it, it crashes and burns when trying to open a screen. |
|
tobing
Member #5,213
November 2004
 |
I could offer some testing: I have Intel ***910GL in the office, and an Intel 82852/82855 GM/GME in my laptop. On both even the OpenLayer demo doesn't work. Well, I thought I might build and debug the demo - but I see that there's no MSVC project file, and quite a few prerequisites. As of now, I don't have SVN or CVS on my PCs (mainly because I'm using Perforce for my own work), so I can only use official releases or snapshots or something. Does any of you have all sources necessary to build OpenLayer 2.0 (or later) and could pack this into a file for me to download? Would make life much easier for me, because I would really like to use OpenLayer in the first place, and help by debugging it, when I can. I guess I would also be able to create a MSVC project file, if that doesn't exist yet. |
|
Richard Phipps
Member #1,632
November 2001
|
Well, I've pointed out the problem with the SVN version with reagrd to the intel chipsets. I had to hack it to get it to work as I don't know enough about the library or OpenGL. It's upto the AllegroGL devs to try to make a proper fix (and not a patch) now in my mind. We've not seen any movement from them yet though (hint!).. EDIT: Tobing, for these problems you only need to use AllegroGL and the examples. (You don't need OpenLayer to see the crash). |
|
tobing
Member #5,213
November 2004
|
I see... even managed to retrieve some projects from SVN and CVS today. Maybe I'll debug into some of the AllegroGL examples to learn some details. A complete source package would be nice nevertheless: I do intend to make use of OpenLayer once it works on my Laptop, and it would be easier if someone could provide a package containing all the sources needed to build it. Of course I can collect all the packages in all the correct version myself, but it takes time to do so, and at least the makers of OpenLayer should have all sources at hand, shouldn't they? It would be nice, really... Edit: I'm trying to build from SVN (using MSVC 7.1 on WinXP), but that does not work: It's missing the generated file 'gl_ext_alias.h' which is created by mkalias.sh, so I cannot do this. Maybe someone could attach this missing file here? Also glx_ext_alias.h and wgl_ext_alias.h, they're also missing. Or someone might confirm that the files contained in the 0.4.0rc4 distribution can be used? Edit: I've built the SVN version successfully, using the generated headers from 0.4.0rc4, and not compiling x.c and djgpp.c - test.c has run fine. Debug. More to come... |
|
Richard Phipps
Member #1,632
November 2001
|
Thanks tobing! |
|
tobing
Member #5,213
November 2004
|
Now I've run all the AllegroGL examples, and had two of them crash, all others run well. Both crashes seems to be totally unrelated and somewhat obscure, and also not related to the Intel graphics chip problem. The bad tests are dumbtest and fonttest... Well, now I'm on my way to build OpenLayer. For this I need to build GlyphKeeper, which took me some moments to get the project configuration right - I couldn't built for target GLYPH_TARGET_ALLEGGL because it didn't compile, only for target GLYPH_TARGET_ALLEGRO. Hopefully that's OK for use with OpenLayer? |
|
Epsi
Member #5,731
April 2005
 |
you can build the latest SVN OL with Freetype and get rid of Glyphkeeper already. ___________________________________ piccolo: "soon all new 2d alegro games will be better. after i finsh my MMRPG. my game will serve as a code reference. so you can understand and grab code from." |
|
tobing
Member #5,213
November 2004
|
Thanks, goods hint. So I built the last Freetype library (2.2.1) and tried to use it with alfont 1.9.2 and GlyphKeeper 0.29 as an exercise: alfont compiles and links fine, but GlyphKeeper does not, because at some place they try to include an internal header file which seems to have changed. So GlyphKeeper remains on Freetype 2.1.10 this way. Now I'll try the latest SVN of OpenLayer. |
|
Evert
Member #794
November 2000
 |
Quote: It's upto the AllegroGL devs to try to make a proper fix (and not a patch)
What is the distinction between 'a proper fix' and 'a patch' in your mind? |
|
tobing
Member #5,213
November 2004
|
OK, I'll give it up for today. I just can't get OL to compile, not the 2.0 distribution, nor the latest SVN version. Oh, I'm using MSVC 7.1 and I'll not use this cbuild that's supplied with OL: I want to use a project file to be able to use all my tools within the IDE. It would be nice if somebody could give me any further information on how to build OL, compiler flags etc. There are many compilation errors, which might be due to a missing #define or some compiler options which have to be set, but so far I haven't found any clue. Thanks in advance! |
|
Richard Phipps
Member #1,632
November 2001
|
I could never get OpenLayer to compile either. Evert: To me, a fix solves the problem, a patch is more of a hack to get it to work. When the Intel chipset problem came up before the first solution was to detect that chipset and not try Multisampling in that case. But now I've seen this problem occur with multiple intel chipset types and so I think a better way is needed than a lot of specialised cases. |
|
Fladimir da Gorf
Member #1,565
October 2001
 |
I've been reported about a compilation problem with the latest SVN version when compiling with GCC, maybe one of the files in the SVN repository hasn't been updated or something. I'll check it out in the evening... Most likely if you're compiling OL for MSVC, I'd blame Microsoft's non-standard compiler for the issues then... OpenLayer has reached a random SVN version number ;) | Online manual | Installation video!| MSVC projects now possible with cmake | Now alvailable as a Dev-C++ Devpack! (Thanks to Kotori) |
|
Evert
Member #794
November 2000
|
Quote: To me, a fix solves the problem, a patch is more of a hack to get it to work. Ok, just understand that when someone says `there's a patch that fixes the issue,' they are likely to mean it in the sense I described. |
|
Richard Phipps
Member #1,632
November 2001
|
|
tobing
Member #5,213
November 2004
|
Thanks for checking. I hope I can continue looking into this tomorrow! Fladimir, I have some years of experience using MSVC and writing portable code, including various Unixes and linux. Usually, the MSVC compiler is quite close to the standard, whereas gcc does compile a lot of things which are not ANSI C++. I can give you examples of this tomorrow (hopefully). Just to ask this: I have access to SVN now, to get files from the depot. Would it also be possible to submit changes? I assume not, so how would that be done? |
|
Bob
Free Market Evangelist
September 2000
|
Quote: To me, a fix solves the problem, a patch is more of a hack to get it to work. Aren't you thinking of a "hack" or a "band-aid"? However, since the bug in question is inside the Intel OpenGL drivers, there is no proper fix AllegroGL can do other than to try not to hit it. This will require keying off the GPU name (which isn't very reliable). -- |
|
Richard Phipps
Member #1,632
November 2001
|
So, how do other libraries such as SDL or Torque2D solve this issue? |
|
Bob
Free Market Evangelist
September 2000
|
They just don't let you use multisampling out of the box -- |
|
Richard Phipps
Member #1,632
November 2001
|
What exactly is multisampling, and should AllegroGL do something similar? |
|
Milan Mimica
Member #3,877
September 2003
|
Richard Phipps said: I thought the latest SVN was supposed to fix that Intel bug? It fixes some of the bugs related to Intel chipsets. More precisely, it fixes bugs in select_pixel_format() reported in logfile. The bug with multisampling is the only one left unfixed AFAIK. I'm not sure what is the proper way to fix it. I'm not even sure it is a Intel bug. The thing is that Intel915 does support multisampling. They are OGL-1.4 compliant. glSampleCoverageARB() does not have to work and "ARB_multisample" does not have to be listed in GL_EXTENSIONS if the extension was promoted to core. Everything is legal on Intel side. glSampleCoverage() works! However, AGL assumes that the ARB version will be available even if extensions was promoted to core. That is true in most cases, but not true enough. The best fix I see is to modify AGL code that manages extensions to use the non-ARB version if the extension was promoted to core. That's not a trivial thing to do.
-- |
|
Bob
Free Market Evangelist
September 2000
|
There are a couple of options available: -- |
|
|
|
