Post MacOSX Torque Game Builder Release Candidate 1 bugs here

#1 I've already reported about the bug with background on MyFishExample demo on Mac Mini. The bug seems to be solved, but unfortunately the TGB works too slow when I am opened MyFishExample demo or game. If you need to do some additional tests, I am ready to assist you.

#2 LevelBuilder, on my PPC Mac mini (OS X 10.4.6), seems to be having a big problem with background images; leading to maxed out processor usage. The only workaround that I've been able to come up with so far is to change the prefs to:
$pref::OpenGL::force16BitTexture = "1"
Obviously, some of the graphics don't look so hot; but at least this gives PPC Mac mini users the ability to run LevelBuilder without significant headaches!

Special thanks to the GG developers who solved the background image corruption and grey particles bugs in LevelBuilder.

#3 More on the PPC Mac mini problem (I'm assuming that the problem is limited to very specific PPC Macs; otherwise lots of people would be noticing it)...

I've been working on a game since Beta1; where it worked flawlessly. After I converted it to Beta4/RC1, the game starts up with a blank white screen that lasts 3-5 seconds; with a brief appearance of the spinning beach ball of death. This all happens during engine initialization. Then the game loads and plays normally.

Since this game doesn't require LevelBuilder, and since I had had problems with Beta2, I simply kept working with Beta1. Today, I went back and converted the game to Beta2; and I again see the long white screen when TGB first starts up.

I've also gone back and tried to duplicate the PPC Mac mini LevelBuilder bugs (described in the threads above) in Beta2. Things have changed a lot, of course, but some of the same large processor hits due to background images do seem to be present.

Something appears to have changed between Beta1 and Beta2/Beta4/RC1; causing these problems.

I wish that I were a good enough C programmer to be able to look at the platformMacCarb source files and provide a solution, but I'm not.

#4 In common/preferences/defaultPrefs.cs::

// Here is where we will do the video device stuff, so it overwrites the defaults
// First set the PCI device variables (yes AGP/PCI-E works too)
initDisplayDeviceInfo();

The engine cannot find this function. Commenting out the line seems to have no affect on TGB.

#5 All my testers, who has mac OS X version lower as 10.4 reports (10.3.9, 10.3.5), that they cannot see anything, only black screen in the full screen mode. Windowed mode is ok.

#6 @ Kostya: The fish demo currently uses a large amount of texture memory, and a full screen transparent particle effect. For comparison, what sort of frame rates are you getting in the fish demo, and what kind of frame rates are you getting in other demos?

@ Don: I'll look into that long startup time. It's probably the resource manager loading in all the textures. We'll see what Shark turns up.

@ Pavel: I'll see if I can reproduce that blank screen bug. Can you give me more information about your tester's Macs? It would help to narrow down the problem.

@ All Mac Mini Users: The PPC Mac Mini does not have very good graphics, with only 32Mb of texture memory. The Intel Mac Mini is even scarier, with a built-in Intel graphics chip, and using system memory. I'm not too surprised that there are speed issues with large textures. We're aware that some of the art resources in the current distribution of TGB are rather large, especially in the fish demo, and we're optimizing some of them for the release.

Thanks for the feedback, all!

This thread is still open, post bugs if you've got 'em.
I'm especially interested in any crashbugs or other show-stoppers.

Share and Enjoy.

/Paul

#7 Paul: it was reported by all my testers who has OS X version 10.3.X, on the 10.4 works ok. And the problem is only on fullscreen mode. We tested one of the original demo from TGB and the same problem, so I hope it is not only in my game.
The version Beta 1 works OK, this problem is in version Beta 4 and RC1.


This is a fragment of a log of one of the Macs with all info:

//-------------------------- 6/6/2006 -- 16:31:14 -----
System & Processor Information:
   MacOS version: 10.3.9
   CarbonLib version: 1.6.0
   Physical RAM: 256MB
   Logical RAM: 2047MB
Unknown Getstalt value for processor type: 0x113
platform layer should be changed to use sysctl() instead of Gestalt() .
   Unknown Processor, assuming x86 Compatible, 1250 Mhz
   FPU detected
   AltiVec detected
 
Math Init:
   Installing Standard C extensions
   Installing Altivec extensions
 
Input Init:

Video Init:
  active displays = 0x3d8080  CGDisplayID = 0x1b01dae4
using display 0x3d8080
   Accelerated OpenGL display device detected.

Initializing chunk mappings...
   o 'TEXT' maps to TextChunk
   o 'SCHK' maps to UnknownChunk
   o 'SCHK' maps to SimChunk
setModPaths: invalid mod path directory name: 'tools'
Loading compiled script common/main.cs.
Missing file: tools/main.cs!
Error: Unable to find specified mod: tools
Loading compiled script game/main.cs.
--------- Parsing Arguments ---------
Error: Unknown command line argument: 

--------- Initializing MOD: Common ---------
Loading compiled script common/preferences/defaultPrefs.cs.
Compiling common/preferences/prefs.cs...
creating path common/preferences/prefs.cs.dso
Loading compiled script common/preferences/prefs.cs.
Loading compiled script common/gameScripts/common.cs.
Loading compiled script common/gameScripts/audio.cs.
Loading compiled script common/gameScripts/canvas.cs.
Loading compiled script common/gameScripts/cursor.cs.
Binding server port to default IP
UDP initialized on port 0
Activating the OpenGL display device...
 OpenGLDevice activating...
OpenGL driver information:
  Vendor: ATI Technologies Inc.
  Renderer: ATI Radeon 9200 OpenGL Engine
  Version: 1.3 ATI-1.3.42
OpenGL Init: Enabled Extensions
  ARB_multitexture (Max Texture Units: 6)
  EXT_blend_color
  EXT_blend_minmax
  EXT_compiled_vertex_array
  EXT_texture_env_combine
  EXT_packed_pixels
  EXT_fog_coord
  ARB_texture_compression
  EXT_texture_compression_s3tc
  (ARB|EXT)_texture_env_add
  EXT_texture_filter_anisotropic (Max anisotropy: 16.000000)
  Vertical Sync
  ATI_FSAA
OpenGL Init: Disabled Extensions
  EXT_paletted_texture
  NV_vertex_array_range
  3DFX_texture_compression_FXT1

 set screen mode 1024 x 768 x 16, fullscreen, force it, dont repaint
Cleaning up the display device...
>> Attempting to change display settings to fullscreen 1024x768x16...
using display 0x3d8080
Dumping list of pixel formats:
 ------------------------------------------------------------
Describing pixel format 0x6351fc
         ALL_RENDERERS    0
           BUFFER_SIZE   16
                 LEVEL    0
                  RGBA    1
          DOUBLEBUFFER    1
                STEREO    0
           AUX_BUFFERS    0
            GREEN_SIZE    5
              RED_SIZE    5
             BLUE_SIZE    5
            ALPHA_SIZE    1
            DEPTH_SIZE   16
          STENCIL_SIZE    0
        ACCUM_RED_SIZE    5
      ACCUM_GREEN_SIZE    5
       ACCUM_BLUE_SIZE    5
      ACCUM_ALPHA_SIZE    0
            PIXEL_SIZE   16
        MINIMUM_POLICY    0
        MAXIMUM_POLICY    0
             OFFSCREEN    0
            FULLSCREEN    1
    SAMPLE_BUFFERS_ARB    0
           SAMPLES_ARB    0
     AUX_DEPTH_STENCIL    0
           COLOR_FLOAT    0
           MULTISAMPLE    0
           SUPERSAMPLE    0
          SAMPLE_ALPHA    0
 ------------------------------------------------------------
Set up AGL windowed support
OpenAL Driver Init:
   Vendor: Any
   Version: OpenAL 1.2
   Renderer: Software
   Extensions:

And the second problem - the RC1 version doesn't read data from ZIP files. Till the Beta 4 it is ok.

Thank you!

#8 Paul, I have to disagree with you about the Mac mini's scant video memory being the problem with the Level Builder.

I've played with this issue a lot; and came up with a simple test... I have a game that is filled with lots of resources and TGB handles it just fine on a PPC Mac mini (the game runs beautifully). It's a typical game, with the various background images declared in .gui files rather than in datablocks. This game loads into the Level Builder without problem; showing all the resources nicely. Add a couple of the background images using the Level Builder's 'Create a New Imagemap' function; and my PPC Mac mini grinds and sputters (especially if you try to place those background images onto the scene). Instead, if you add those same background images using datablocks in script, the Level Builder runs without problems (there's no issue with creating a level, complete with layered background images and filled with sprites and particle effects). This simple test would seem to indicate that the problem isn't my computer, but something in the Level Builder.

I've gone through the Fish Demo and converted the background images from 1600x1200 to 800x600 and get the same (poor) results when running the Level Builder. A couple layered background images, a particle effect and a single fish shouldn't make the Level Builder run like a dog and scramble the background images.

On another subject... Since this thread is still open, does that mean that Mac users won't be getting RC2 anytime soon?

#9 Paul, I ran Shark on that game of mine that has the long white screen at startup (5-6 seconds). Shark is nearly incomprehensible to me; so if you'd like me to send you any reports, just let me know.

Things that seemed significant while running Shark from the time the game launches to when the mainMenuGui loads::

Time Profile

GLEngine --- gleExtractPixel_RGBA_Ubyte --- 22% in RC1 and 6% in Beta1
GLEngine --- gleTextureImagePut_Slow --- 20% in RC1 and 5% in Beta1
TGB --- 10% (total of lots of entries)

Malloc Trace (seemed fairly normal, to me)

TGB --- 48%
CoreFoundation --- 17%

Static Analysis of TGB (haven't a clue!!!)

XviD_Codec yv12_to_yuyv_altivec_unaligned_c --- 10%
XviD_Codec yv12_to_uyvy_altivec_unaligned_c --- 10%

[The game in Beta1 does not exhibit the white screen at startup I now see in RC1.]

#10 @Don: Please send me a zip of an example TGB project, that has one level that runs dog-slow, and one that runs well. I need to dig in & see what *exactly* the differences are. On the Shark stuff, thanks for the feedback, I'll keep an eye out for those.

#11 Paul... I've tried duplicating the Level Builder problems (with my PPC Mac mini) using very small projects, but am having trouble recreating the exact situations that made me comment. I'll keep working on it and send you relevant files if I can come up with a scenario that is easily reproducible.

I ran Shark on Level Builder (with simple projects); and consistently come up with the following bottleneck:
~80% --- com.apple.ATIRadeon8500 --- ATIRadeon8500::submit_buffer(unsigned long*, unsigned long, unsigned long)

Level Builder just doesn't like my graphics card (or there is some kind of conflict between the Level Builder code and ATIRadeon8500.kext)! It's a good thing that TGB doesn't have similar problems.

Oh well, the workaround of setting force16BitTexture=1 in Prefs allows me to use Level Builder.

#12 Something that gets rid of the "background picture scramble" in fish demo is setting and re-setting my PPC mac-mini's monitor resolution.

These are not show-stopper bugs, but some might be easy fixes:

In general, mac users use cmnd-s to save, not ctrl-s.

In general, there are 2 delete keys on the mac keyboard. One of them is actually the backspace. Mac users are used to using the backspace key to delete selected items.

T2D's menu bar is hidden under Finder's window bar in windowed mode, on 800x600 resolution . Re-size/reposition window option? When T2D's menu bar is accessable, anything at the very bottom (like down arrows) is not.

The graph editor for the TGB particle system--key points and lines between them are not always displayed. Click and drag doesn't work. Clicking the "forward" button, and then "back," displays the keypoints and lines between them.

#13 Thank you Jeffrey for coming up with the workaround for the scrambled images in the Level Editor. It certainly helps the situation!

Unfortunately, it doesn't seem to help the extraordinary processor usage generated by the Level Editor:
83.3% --- com.apple.ATIRadeon8500 --- ATIRadeon8500::submit_buffer(unsigned long*, unsigned long, unsigned long)
often (but not always!) shows up as a problem when I run Shark on the Level Editor.

Hopefully this will help to identify the bug in the Level Editor. [Yes, I think that it is caused by a bug; and not because Mac minis are too underpowered to handle the job. The fact that the engine works wonderfully on my PPC Mac mini, but the Level Editor doesn't would seem to indicate a bug. Perhaps I'm wrong; but the situation just seems a little odd. Now that changing the screen's resolution suddenly makes one of the problems go away would also tend to support the notion that a bug exists.]

#14 Preface: I just bought T2D, and haven't gone through much of a tutorial nature yet. I also don't know which of these bugs are Mac-specific and which are PC-specific.

Little minor things that should be cleaned up in the Mac version of T2D editor:

- Move the mouse over the window's title bar and click. The mouse's copy that's inside the window will open a menu. This shouldn't happen -- the mouse is in the title bar, not the menu.

- Looping sound effects that are playing when you quit a program that was run using "Run Game" don't stop when you return to the editor.

- Likewise, the editor doesn't restore its original window size, if "Fullscreen" was selected while the game was being run.

- You can't resize the editor's window. I'd like to have it close to full-screen, but with the normal system mouse showing up outside the window. Doesn't seem to be possible. The rest of my comments in this list are rather petty... However, not being able to resize the window is a BIG drawback.

- There are times when the cursor shown in the editor's window isn't really where the cursor "really" is. For instance, when you first start up, it's shown in the upper-left corner. As soon as you move the mouse, it appears wherever it *reallly* is, which is often a huge jump. It's happened to me other times too, but I don't remember when... perhaps when "Run Game" ended?


- Whenever you have to type text in somewhere (i.e. New Project), try clicking somewhere to move the cursor so you can add/delete text at that location, and the entire thing highlights (even though there is still a blinking cursor in the middle of it), and typing anything deletes the entire thing instantly. Strangely enough, if you click and drag the cursor slightly at the place you want to insert text, it'll unhighlight the text! A single click, without the drag, should do this.

- Any open window with a close box should also have the keyboard shortcut Command-W.

- Likewise, any dialog with cancel should allow Command-period to cancel. (In some cases you may also want to enable the return key to accept, but this should not be the case if the return key has other uses in the dialog box, like adding a carriage return to a block of text.)

- If you want to get really technical, all dialog boxes (i.e. New Project and similar) should change somehow when TGB goes into the background. Blinking cursors should stop blinking, and ideally the title bar would change (or perhaps the whole dialog "dim out" somehow) to indicate that the dailog box is inactive. Right now I'm typing this note into TextEdit, but I have a New Project dialog box open in TGB, and the cursor is blinking in the text field, making it LOOK like any text I type should go in that dialog, where in reality that dialog is in the background and won't accept any text. It just doesn't look like it. [Okay, I admit this is picky, but hey, I'm a Mac user. What do you expect?]

- The Scene View pop-up menu should not have "Remove this View" option (it should be dimmed out) if there is only 1 view. It does nothing if selected anyway...

- I used the Help menu to "View Keyboard Shortcuts", and clicked on the Key Bindings tab, then I dismissed the dialog somehow without changing anything, and the editor jumped from normal size to full-screen! (Making my system cursor disappear when the mouse is moved outside the window.) What happened here???

Worse yet, my previous trick to get the editor out of fullscreen mode is no longer working. Previously, I would run a demo using Run Game, choose Options, and uncheck the Fullscreen box. Then I'd quit the demo, and viola, the editor would be back to normal. Not anymore! In fact, now when I run a demo, the menu items at the bottom of the screen are *under* my MacOS X dock, making it very difficult to click on them. And when I tried to check and uncheck the fullscreen box, the window didn't shrink, but instead the title bar moved up higher, so it was under the MacOS X menu, so I was unable to drag the window around anymore, not making it impossible to get access to any buttons that are covered up by the OS X dock.

- Is there any way to scroll around the Scene view? If so, it's not obvious... There are no scroll bars, and pushing Control, Option, or Command and dragging inside the scene without the mouse don't turn it into a hand and result in dragging... Granted, I haven't read the documentation yet, but you shouldn't have to for something as simple as scrolling. If it can't be figured out, it's not intuitive the way it's done.

#15 I also experience no fullscreen mode with OSX 10.3.8. I have a "stock" Mac Mini w/512mb ram. Fullscreen mode just shows a black window.

RC1 reads zip files ok for me.

#16 Huh, I could've sworn I had posted this last night. But now I don't see it, so I'm re-posting. This is a crash log. That's right, Torque 2D's editor crashed on me -- the very first day I was using it too.

I had opened the Example Fish Game, and had messed around with editing a sprite's collision polygon, then I tried opening the Example Fish demo (not the fish game, but the fish demo). It had the standard pause for opening a new project, and then crashed.

The log is too long for this forum, so I've trimmed it.




Date/Time: 2006-06-13 00:25:30.437 -0700
OS Version: 10.4.6 (Build 8I1119)
Report Version: 4

Command: Torque Game Builder
Path: /Developer/TorqueGameBuilder/games/Torque Game Builder.app/Contents/MacOS/Torque Game Builder
Parent: WindowServer [65]

Version: Copyright 2001-2006 GarageGames.com, Inc. All rights reserved. (1.1)

PID: 324
Thread: 0

Exception: EXC_BAD_ACCESS (0x0001)
Codes: KERN_PROTECTION_FAILURE (0x0002) at 0x000000c8

Thread 0 Crashed:
0 ...agegames.torqueGameBuidlder 0x001d847c 0x1000 + 1930364
1 ...agegames.torqueGameBuidlder 0x00197a0f 0x1000 + 1665551
2 ...agegames.torqueGameBuidlder 0x00198ce4 0x1000 + 1670372
3 ...agegames.torqueGameBuidlder 0x00198f22 0x1000 + 1670946
4 ...agegames.torqueGameBuidlder 0x0019e396 0x1000 + 1692566
5 ...agegames.torqueGameBuidlder 0x0019e743 0x1000 + 1693507
6 ...agegames.torqueGameBuidlder 0x0006b77b 0x1000 + 436091
7 ...agegames.torqueGameBuidlder 0x00069a19 0x1000 + 428569
8 ...agegames.torqueGameBuidlder 0x00069a19 0x1000 + 428569
9 ...agegames.torqueGameBuidlder 0x00069a19 0x1000 + 428569
10 ...agegames.torqueGameBuidlder 0x00069a19 0x1000 + 428569
11 ...agegames.torqueGameBuidlder 0x00069a19 0x1000 + 428569
12 ...agegames.torqueGameBuidlder 0x00069a19 0x1000 + 428569
13 ...agegames.torqueGameBuidlder 0x00074a5a 0x1000 + 473690
14 ...agegames.torqueGameBuidlder 0x0006e2d5 0x1000 + 447189
15 ...agegames.torqueGameBuidlder 0x0006e3b8 0x1000 + 447416
16 ...agegames.torqueGameBuidlder 0x002277d8 0x1000 + 2254808
17 ...agegames.torqueGameBuidlder 0x0006b6cb 0x1000 + 435915
18 ...agegames.torqueGameBuidlder 0x00074a5a 0x1000 + 473690
19 ...agegames.torqueGameBuidlder 0x0006e2d5 0x1000 + 447189
20 ...agegames.torqueGameBuidlder 0x0006e3b8 0x1000 + 447416
21 ...agegames.torqueGameBuidlder 0x000d8aad 0x1000 + 883373
22 ...agegames.torqueGameBuidlder 0x000fef42 0x1000 + 1040194
23 ...agegames.torqueGameBuidlder 0x000ffbe1 0x1000 + 1043425
24 ...agegames.torqueGameBuidlder 0x000bafc6 0x1000 + 761798
25 ...agegames.torqueGameBuidlder 0x001320cf 0x1000 + 1249487
26 ...agegames.torqueGameBuidlder 0x000bb77f 0x1000 + 763775
27 ...agegames.torqueGameBuidlder 0x001451e4 0x1000 + 1327588
28 com.apple.HIToolbox 0x92f317ce TimerVector + 31
29 com.apple.CoreFoundation 0x90823bc9 CFRunLoopRunSpecific + 3341
30 com.apple.CoreFoundation 0x90822eb5 CFRunLoopRunInMode + 61
31 com.apple.HIToolbox 0x92ef9c5c RunCurrentEventLoopInMode + 285
32 com.apple.HIToolbox 0x92ef929a ReceiveNextEventCommon + 184
33 com.apple.HIToolbox 0x92f418b5 _AcquireNextEvent + 58
34 com.apple.HIToolbox 0x92f4170c RunApplicationEventLoop + 150
35 ...agegames.torqueGameBuidlder 0x001431f4 0x1000 + 1319412
36 ...agegames.torqueGameBuidlder 0x00002e3e 0x1000 + 7742
37 ...agegames.torqueGameBuidlder 0x00002d59 0x1000 + 7513

Thread 1:
0 libSystem.B.dylib 0x90037e07 mach_wait_until + 7
1 libSystem.B.dylib 0x9003a572 usleep + 82
2 ...agegames.torqueGameBuidlder 0x00133f3c 0x1000 + 1257276
3 ...agegames.torqueGameBuidlder 0x00133f9c 0x1000 + 1257372
4 libSystem.B.dylib 0x90024a27 _pthread_body + 84

Thread 2:
0 libSystem.B.dylib 0x9000a4e7 mach_msg_trap + 7
1 com.apple.CoreFoundation 0x9082369a CFRunLoopRunSpecific + 2014
2 com.apple.CoreFoundation 0x90822eb5 CFRunLoopRunInMode + 61
3 com.apple.audio.CoreAudio 0x914549f2 HALRunLoop::OwnThread(void*) + 158
4 com.apple.audio.CoreAudio 0x9145480d CAPThread::Entry(CAPThread*) + 93
5 libSystem.B.dylib 0x90024a27 _pthread_body + 84

Thread 3:
0 libSystem.B.dylib 0x90049207 semaphore_timedwait_signal_trap + 7
1 com.apple.audio.CoreAudio 0x91461af8 CAGuard::WaitFor(unsigned long long) + 212
2 com.apple.audio.CoreAudio 0x91461a1a CAGuard::WaitUntil(unsigned long long) + 66
3 com.apple.audio.CoreAudio 0x9146048a HP_IOThread::WorkLoop() + 690
4 com.apple.audio.CoreAudio 0x914601d1 HP_IOThread::ThreadEntry(HP_IOThread*) + 17
5 com.apple.audio.CoreAudio 0x9145480d CAPThread::Entry(CAPThread*) + 93
6 libSystem.B.dylib 0x90024a27 _pthread_body + 84

Thread 0 crashed with i386 Thread State:
eax: 0x000000c8 ebx: 0x0000000e ecx:0x00000000 edx: 0xbf7f5c92
edi: 0x01b5c800 esi: 0xbfffdd80 ebp:0xbfffdd58 esp: 0xbfffdce0
ss: 0x0000002f efl: 0x00010246 eip:0x001d847c cs: 0x00000027
ds: 0x0000002f es: 0x0000002f fs:0x00000000 gs: 0x00000037

Model: iMac4,1, BootROM IM41.0055.B03, 2 processors, Intel Core Duo, 2 GHz, 1 GB
Graphics: ATI Radeon X1600, ATY,RadeonX1600, PCIe, 128 MB
Memory Module: DIMM1/BANK 1, 1 GB, DDR2 SDRAM, 667 MHz
AirPort: spairport_wireless_card_type_airport_extreme (0x14E4, 0x89), 103.2 (3.120.28.3)
Bluetooth: Version 1.7.3f4, 2 service, 1 devices, 1 incoming serial ports
Network Service: Built-in Ethernet, Ethernet, en0
Serial ATA Device: WDC WD2500JS-40NGB2, 232.89 GB
Parallel ATA Device: MATSHITADVD-R UJ-846, 552.38 MB
USB Device: Built-in iSight, Micron, Up to 480 Mb/sec, 500 mA
USB Device: Macally iKey, Macally, Up to 12 Mb/sec, 500 mA
USB Device: Macally iKey, Macally, Up to 12 Mb/sec, 100 mA
USB Device: USB Mouse, Logitech, Up to 1.5 Mb/sec, 500 mA
USB Device: Bluetooth HCI, Up to 12 Mb/sec, 500 mA
USB Device: IR Receiver, Apple Computer, Inc., Up to 12 Mb/sec, 500 mA

#17 @Vern: thanks for the crashlog, and for the suggestions. Unfortunately, I can't do a thing *with* that crashlog, because it's not got any debugging symbols in it. Please compile the debug build of the engine, from source, and try to reproduce the issue, then post the stack trace. Thanks!