Mantis - ISSUES

0000074: Star Wars Episode I Racer

lxdream sees otherwise working game as an audio file, upon load it goes into playback mode.

0000073: mercurial version fails to compile due to undefined G_DIR_SEPARATOR

This bug affects the current revision (1285) of the mercurial repo. It also happens for me with the 0.9.1 source avalable from
http://www.lxdream.org/count.php?file=lxdream-0.9.1.tar.gz [^]

When I run './configure' then 'make', the compilation aborts due to the following errors:

---
lxpaths.c: In function ‘get_absolute_path’:
lxpaths.c:116:47: error: ‘G_DIR_SEPARATOR’ undeclared (first use in this function)
lxpaths.c:116:47: note: each undeclared identifier is reported only once for each function it appears in
lxpaths.c: In function ‘get_filename_at’:
lxpaths.c:129:53: error: ‘G_DIR_SEPARATOR’ undeclared (first use in this function)
---

and after fixing that file, same thing happens elsewhere:

---
drivers/cdrom/cd_gdi.c: In function ‘gdi_image_read_toc’:
drivers/cdrom/cd_gdi.c:76:5: warning: implicit declaration of function ‘g_path_get_dirname’ [-Wimplicit-function-declaration]
drivers/cdrom/cd_gdi.c:76:13: warning: assignment makes pointer from integer without a cast [enabled by default]
drivers/cdrom/cd_gdi.c:126:67: error: ‘G_DIR_SEPARATOR’ undeclared (first use in this function)
drivers/cdrom/cd_gdi.c:126:67: note: each undeclared identifier is reported only once for each function it appears in
---

I am running Ubuntu 12.04 (64bit). For me G_DIR_SEPARATOR is defined in the file /usr/include/glib-2.0/glib/gfileutils.h

I am no coder, but my solution was to add the line
#include
at the beginning of the files lxpaths.c and drivers/cdrom/cd_gdi.c.

Also, glib.h seems to include all other glib/g*.h files, including glib/gfileutils.h. So an alternative solution is to replace the lines
#include
#include
at the beginning of the files lxpaths.c and drivers/cdrom/cd_gdi.c with simply:
#include

I don't know which is better. I tried both solutions and they both work. It's probably a question of style. As I said, I am no coder :-) I have attached a patch that implements the former solution (as it looked more minimalistic).

Cheers,

Malte

0000072: Featur request for GUI: Screenshots

It would be nice if lxdream had the feature to make a screenshot of the game by pressing a button, preferably in PNG format. I suggest a function key, e.g. F12 as default.

0000071: Add support for pkg-config to libgl and libisofs

Hello,
I've made a patch to detect automatically (via pkg-config) libgl and libisofs.
I don't know about linux distro but my FreeBSD have the required .pc files.

0000070: No Sound + Infinite Loop

I'm using the oggvorbis library, and it works on the Dreamcast but doesn't work on lxDream. After a while, it will go into an infinite loop.

0000069: sh4/sh4trans.c:238: xlat_get_code_by_vma: Assertion `(sh4_icache.page_vma == ((sh4r.pc) & sh4_icache.mask))' failed.

I've tried this on 0.9.1 and 0.8.4.

I'd like to preface this bug by saying that it could be caused by an error on my end, but I can't find one. Also, my program works on the hardware, but crashes when run on the emulator, giving me this error:

sh4/sh4trans.c:238: xlat_get_code_by_vma: Assertion `(sh4_icache.page_vma == ((sh4r.pc) & sh4_icache.mask))' failed.

Here's the code
//Main.cpp
...
//If this if statment is commented out, the program runs fine
if (levels[gameState.getCurrentLevel()].levelComplete(enemies))
gameState.nextLevel();
...

And the functions:
int GameState::getCurrentLevel() const
{
return currentLevel;
}

bool Level::levelComplete(std::vector &enemies)
{
for (std::vector::iterator it = enemies.begin();
it != enemies.end(); ++it)
{
if (!it->getDead())
return false;
}

return true;
}

I saw a similar bug already posted that was resolved. Was that resolved in a newer version than I'm testing? If it's not a bother, could you update me about this bug via email: EnigmaticCoder [atttt] hotmail [dotttt] com

If you need any more code, I'd be happy to send it to you.

0000068: sh4_finalize_instruction: Assertion `(sh4_icache.page_vma == ((sh4r.pc) & sh4_icache.mask))' failed.

Ran into this while trying a test:
gdb --args ./lxdream /usr/local/dc/kos/examples/dreamcast/hello/hello.elf
GNU gdb 6.8-debian
Copyright (C) 2008 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html> [^]
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "x86_64-linux-gnu"...
(gdb) r
Starting program: /home/harley/Downloads/lxdream/src/lxdream /usr/local/dc/kos/examples/dreamcast/hello/hello.elf
[Thread debugging using libthread_db enabled]
[New Thread 0x7f877b7e5810 (LWP 8389)]
[New Thread 0x7f8766f42950 (LWP 8399)]
[Thread 0x7f8766f42950 (LWP 8399) exited]
[New Thread 0x7f8766f42950 (LWP 8400)]
[New Thread 0x7f8766741950 (LWP 8401)]
10:24:32 A0000000 WARN Unable to load file 'IP.BIN': No such file or directory

--
KallistiOS ##version##: Wed May 13 13:39:27 CDT 2009
harley@little-jack:/usr/local/dc/kos
thd: pre-emption enabled, HZ=100
10:24:32 8C021B6E WARN ARM Halted: BRANCH to invalid address FFFFFFE8 at 00000004
maple: active drivers:
Dreameye (Camera): Camera
Sound Input Peripheral: Microphone
PuruPuru (Vibration) Pack: JumpPack
VMU Driver: Clock, LCD, MemoryCard
Mouse Driver: Mouse
Keyboard Driver: Keyboard
Controller Driver: Controller
DMA Buffer at ac0923c0
vid_set_mode: 640x480IL NTSC
fs_romdisk: mounting image at 0x8c0489d4 at /rd
dc-load console support enabled
maple: attached devices:
A0: Dreamcast Controller (01000000: Controller)

Hello world!

fs_iso9660: disc change detected
arch: shutting down kernel
net_unreg_device: 'bba' isn't registered
maple: final stats -- device count = 1, vbl_cntr = 4, dma_cntr = 4
vid_set_mode: 640x480IL NTSC
fs_romdisk: unmounting image at 0x8c0489d4 from /rd
max system bytes = 147908
system bytes = 147908
in use bytes = 5852
KM_DBG: Still-allocated memory blocks:
INUSE 8c0ac578: size 12, thread 1, addr 8c0104d8, type malloc
INUSE 8c0abf70: size 1024, thread 1, addr 8c03f594, type malloc
INUSE 8c091280: size 2, thread 1, addr 8c040a82, type malloc
INUSE 8c090be8: size 12, thread 1, addr 8c0104d8, type malloc
INUSE 8c0909c0: size 12, thread 1, addr 8c02dd94, type malloc
INUSE 8c08ebf0: size 12, thread 1, addr 8c01e8f2, type malloc
INUSE 8c08e5e8: size 1024, thread 1, addr 8c03f594, type malloc
lxdream: ./sh4/sh4core.in:242: sh4_finalize_instruction: Assertion `(sh4_icache.page_vma == ((sh4r.pc) & sh4_icache.mask))' failed.

Program received signal SIGABRT, Aborted.
[Switching to Thread 0x7f877b7e5810 (LWP 8389)]
0x00007f8777b27fb5 in raise () from /lib/libc.so.6
(gdb) bt
#0 0x00007f8777b27fb5 in raise () from /lib/libc.so.6
0000001 0x00007f8777b29bc3 in abort () from /lib/libc.so.6
0000002 0x00007f8777b20f09 in __assert_fail () from /lib/libc.so.6
0000003 0x000000000041d8a8 in sh4_finalize_instruction () at ./sh4/sh4core.in:242
0000004 0x00000000004189b5 in sh4_core_exit (exit_code=2) at sh4/sh4.c:306
0000005 0x0000000000416f2e in dreamcast_stop () at dreamcast.c:255
0000006 0x0000000000410793 in dcload_syscall (syscall_id=8389) at dcload.c:151
0000007 0x00000000004509fd in sh4_translate_run_slice (nanosecs=1000000) at sh4/sh4trans.c:50
0000008 0x0000000000418b39 in sh4_run_slice (nanosecs=1000000) at sh4/sh4.c:271
0000009 0x000000000041728f in dreamcast_run () at dreamcast.c:233
0000010 0x0000000000456cfd in gui_main_loop (run=1) at gtkui/gtkui.c:251
0000011 0x000000000040c9f0 in main (argc=2, argv=0x7fff8392de68) at main.c:270

and here's the code (don't laugh since I was just doing a quick test) that causes the crash:
cat hello.c
#include
extern uint8 romdisk[];
KOS_INIT_FLAGS(INIT_DEFAULT | INIT_MALLOCSTATS);
KOS_INIT_ROMDISK(romdisk);

int main(int argc, char **argv) {
printf("\nHello world!\n\n");
file_t cd = fs_open("/cd", O_DIR);
if(!cd)
return 1;

dirent_t *cddir;
while((cddir = fs_readdir(cd)))
printf("%s\n", cddir->name);

return 0;
}

I'm not currently able to test it on real hardware, but it /should/ work. When I get home from work tonight I'll give it a try. Attached is the crash log in case this butchers the formatting of the code and backtrace.

0000067: cdi bug x86_64 OS X

Hi Nathan,

I tried compiling lxdream with -arch x86_64 on OS X and I came across a bug with the cdi image loading code.

in cdi.c

The code to seek to the header_offset is broken for CDI_V35_ID

You need to cast to long when using a negative offset.

I changed the code to

if( trail.cdi_version == CDI_V35_ID ) {
fseek( f, -(long)trail.header_offset, SEEK_END );
} else {
fseek( f, trail.header_offset, SEEK_SET );
}

which seemed to get it working again.

Regards,
Eric

0000066: OS X Path properties panel

The OS X path properties panel is really poor. Needs to at least use file choosers, and give the user immediate feedback as to whether the files they've selected are actually valid.

0000065: texcache invariant failure

As reported by sky:
Same configuration, with Sega Extrem Sports. It happen on game loading, once on two.

Code:

18:24:50 8C00B8FA WARN Attempted byte read to missing page: A0600004
Active list: 205, 215, 214, 213, -1, -1, -1, 189, 204, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 181, 201, -1, 166, -1, -1, -1, 241, -1, -1, -1, -1, -1, -1, -1, -1, 215, -1, -1, -1, -1, -1, -1, 206, -1, 116, -1, 249, -1, 212, -1, -1, -1, -1, -1, -1, -1, 203, -1, 175, 207, -1, -1, -1, -1, -1, -1, -1, -1, -1, 162, -1, -1, -1, -1, -1, 211, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, 239, -1, -1, -1, -1, -1, 230, 167, -1, -1, -1, 115, -1, -1, 224, 148, 174, 187, 208, -1, -1, -1, -1, -1, 169, 253, -1, -1, -1, -1, -1, -1, -1, -1, 135, 211, 200, 202, 209, 195, -1, -1, -1, -1, -1, -1, -1, -1, -1, 192, -1, -1, -1, 177, -1, -1, 210, -1, 243, -1, -1, -1, -1, -1, 167, -1, -1, 164, -1, -1, -1, -1,
pvr2/texcache.c:576: failed assertion `next != slot'

Program received signal SIGABRT, Aborted.
0x9003d66c in kill ()
(gdb) bt
#0 0x9003d66c in kill ()
0000001 0x9010e8cf in raise ()
0000002 0x9010d422 in abort ()
0000003 0x0005cebd in __eprintf ()
0000004 0x000332cc in texcache_get_texture ()
0000005 0x00035cd4 in pvr2_scene_load_textures ()
0000006 0x00036360 in pvr2_scene_render ()
0000007 0x0002d8d5 in mmio_region_PVR2_write ()
0000008 0x0e3729f3 in ?? ()
Previous frame inner to this frame (gdb could not unwind past this frame)

0000032: Implement shader programs (GLSL)

PVR2 seems to do colour clamping after computing the final fragment colour, unlike the GL fixed-function pipeline which clamps each component separately prior to interpolation. Note that we're likely to need this for correct fogging as well.
Also as a bonus, after this we can rip rendbkg.c out almost completely.

1. Add test cases for PVR2 behaviour. At least Tex modulate and Quad tests needed to confirm the above.
2. Add simple shader programs as above
3. Remove obsolete hacks

Fallback: In the absense of GLSL support, system should run normally using the fixed-function pipeline - it will just be less accurate.

0000008: Implement render-to-texture

Currently broken.

0000061: OpenBSD Ports

I have trying to ports lxdream to OpenBSD. For the moment it's just compile fine. But i haven't try to compile SVN. GLX and esd don't run the emulator is unusable for the moment

0000064: OS X 64-bit build crashes in translator

Probably pointer sloppiness - OS X organizes memory a little differently

0000063: Add OS X Quick Lock (preview) + Spotlight support for save states

ie finder integration. Nice to have, but probably a bit of work involved.

0000062: Add OS X HID (joystick etc) supporrt

Probably easy enough to do, just needs the hardware interface

0000044: Mac OS X Port

Primarily need the CD driver and audio driver - a native video driver is probably not viable without a native UI, and that's going to take too long for the current release.

0000060: Crash in sh4 translation

Reported by Arthonis with Ready To Rumble 2. No other way known to reproduce as yet. Backtrace:

0000004 0x08075268 in xlat_promote_to_temp_space (block=0xb3fa7fec)
at sh4/xltcache.c:373
0000005 0x08075515 in xlat_extend_block (newSize=1303511014) at sh4/xltcache.c:433
0000006 0x080aa5cf in sh4_translate_basic_block (start=2351005696)
at sh4/sh4trans.c:174
0000007 0x080aa321 in sh4_xlat_run_slice (nanosecs=1000000) at sh4/sh4trans.c:91
0000008 0x0805bee2 in dreamcast_run () at dreamcast.c:194

0000051: OSMesa (software rendering) support

For people with intel and the like cards that are never going to work w/ hardware acceleration

0000029: Frustum clipping bug

Polygons subject to view frustum clipping appear to have badly distorted texture coordinates wrt clipped vertexes (colour coordinates are probably equally distorted. Opengl information suggests that the problem may be caused by non-planar polygons - check if the 1/z is correct?

Otherwise may need to do our own clipping, which is annoying but doable.

0000057: Add configure test for framebuffer support

It's possible some older libGLs won't actually have the fbo entry points - add configure test to disable FBO support when missing. As reported by losinggeneration in http://www.lxdream.org/forums/viewtopic.php?t=104 [^]

0000059: Finish handling of all CD read modes

Some testing required as well.

0000058: Add DSF program support

Looks to be a thin wrapper around the ARM7 binary image. See http://www.zophar.net/tech/files/psf_format15.txt [^] .

0000056: KOS s3mplay broken

looks like it hasn't worked for a long time :(

0000055: Fix relative path handling for open/save dialogs

Currently if the configured path is relative, the dialog will complain about it:

libgnomevfs-CRITICAL **: gnome_vfs_get_uri_from_local_path: assertion `g_path_is_absolute (local_full_path)' failed

probably need to convert relative paths first before it gets to that point.