Fail To Make: "Bus Error"

[tiga]

Hi there! I have some problems compiling my program for the DS, so what I'm up to is to ask if anyone can give me a hint to solve this. The problem looks like this:
So, I was able to compile my prog all the time, having done my very own 3D engine made with Xcode and SDL. Had also done everything ready to just compile it via devkit without changing anything. This just worked well, except for one thing: All polygons (triangles) were drawn awful wrong. Loading the map and collision detections works just like on my computer, but no matter if I use the BSP-Tree for drawing or just draw every single polygon, it get screwed up (say, it's just a bunch of triangles -- they look somehow a bit like they should be, just WAY smaller, displaced, and a bit 'smashed'... hard to explain). BTW: I'm not converting the float values myself, I still using the "glVertex3f" methods. But as far as I tested, converting all those floats into v16 values doesn't change anything...

So, I thought "Hmmm... My devkitPro version is 'quite' old (r21), there must be a newer one, that might fix this problem!" -- so thought, so done. I dowloaded everything (newest versions) told in the "getting started" section at the wiki, placed everything where it should be -- resulting in being not able to compile my program anymore. When calling the "make" command in the terminal, I can see the compiler compiling everything needed to link the DS ROM succesfully, but when it comes to linking itself, the compiler gets an error. This is what I get:

Code: Select all

linking DSTEST.elf
built ... DSTEST.arm9

make[1]: *** [/Users/mensk/Desktop/DSTEST/DSTEST.nds] Bus error
make: *** [build] Error 2

Stupid as I am, I deleted the older devkit version, and exactly this version seems to be not available anymore...
Does anyone have a suggestion? Would a newer version possibly fix my first problem? If not, what could cause this happening? And how may I bring my devkit back to life? :'(

I'm running a G5 (PPC) with MacOS 10.4.11 (Tiger).
All DS testing was done on a real DS.

I'll very appreciate anyone's help!!

Oh and guys: YOU ARE AWESOME!! Thank you for devkitPro!!

[DarkShadow44]

If your free space for E-Mails is bigger than 8 mb , I can send you devkitArm r21... (Of course I would need your adress)
Or can I use attchments ?

But you should try to re-install devkitpro (I've also the newest and no problems), perhaps after a backup.

[tiga]

Oh, that would be so great! Sent you a PM.

Could you tell me how you did that? As I said, just downloaded it, unpacked it, placed it where it should be -- DONE!
Also, it really looks like it WOULD work, except for the last few actions to complete the progress...

I dunno if that problem is caused by the Makefile, but I doubt so, because I can't even make all those nice examples...

Well, anyway, thank you kindly!

[WinterMute]

If your free space for E-Mails is bigger than 8 mb , I can send you devkitArm r21... (Of course I would need your adress)
Or can I use attchments ?

Please do *not* distribute devkitPro issued binaries by any means. This causes major problems with support issues.

[DarkShadow44]

Did you only re-install devkitARM ?
Then try to renome your "devkitpro" folder, and create a new one.
Then download "devkitProUpdater-1.5.0.exe" and install ALL again.

First try to download the new examples, these should be compiled.

[tiga]

Okay, it seems like you misunderstood my situation:
I'm using MacOS X! Not Windows! There is not yet an installer/updater for other systems than Windows. That's why I need to install it myself.
So, your kindly offer to send me the version I had before probably would have been useless anyway. But thank you for your effort!

Has anyone else experienced this problem before? Could the version r21 have caused this error when drawing the polygons (as you ought him to not distribute issued binaries)?
Is there absolutely nothing I can do to fix this problem?

Oh, and BTW: It was possible to draw polygons by hardcoded values... this happened just by drawing using variables stored in vertices stored in arrays created with 'new'... (tried it too by malloc them, still didn't work).

*sigh* Everything went so well...

[WinterMute]

See if either of these binaries work for you.

https://sourceforge.net/project/showfil ... _id=683164

[tiga]

Hmmm... no, it doesn't work. Neither of them. Everything's just like before: "make" command-> starting to build all necessary files to link the DS ROM file-> when it comes to linking, it fails.

BTW: I came to the awesome ingenious idea to install it on my very lame 600mhz Windows PC, and it works (newest version).
But I still don't get the problem with that drawing. No matter what I try, it's always the same results.

All polygons do have pointers to vertices stored in an array. And every BSP tree has a vector containing pointers to those polygons, also stored in an array. As said, no matter if just every polygon in the array is drawn or the BSP trees are used... that doesn't change anything...

Everything seems correct (by printing and by collision detection), it's just the drawing that fails -- even the camera positioning works just like on my mac/pc... I'm starting to yield to despair...

Anyway, thank you for helping!

EDIT:

Oh, and as you're mentioning it, when using version r24 (and I think r25 too), there's another line appearing when making the file. It's an ndstool command, and looks like this:

Code: Select all

ndstool -c /Users/mensk/Desktop/DSTEST/DSTEST.nds -9 /Users/mensk/Desktop/DSTEST/DSTEST.arm9 -b /opt/local/devkitPro/libnds/icon.bmp "DSTEST;www.devkitpro.org;www.drunkencoders.com" 
make[1]: *** [/Users/mensk/Desktop/DSTEST/DSTEST.nds] Bus error
make: *** [build] Error 2

[WinterMute]

I uploaded the ppc ndstool binary from r21 at the same url https://sourceforge.net/project/showfil ... _id=683164

That should get you going in the meantime.

It would be useful if you could run the universal binary through gdb & give me a backtrace if that's possible. The binary works fine for me through Rosetta here, not entirely sure what's going on.

[tiga]

Yes, you're right -- that works! I'm expecting now that It's not an installation failure due to I'm able to succesfully compile using the PPC binary.

Uhm, to be honest, I don't know exactly how to give you a backtrace... :-/ I'm quite stupid right now as I don't know what you mean... I'm sorry!
But I would like to place myself at your disposal, to get rid of this problem for myself an all the other MacOS X PPC users out there using devkitARM.