Reported in 0.37b14.1:
- Add -steadykey support from MAME.

Reported in 0.37b15.1:
- Certain games, such as kabuki, hang when running xmame.xgl, unless you first
  run another version of xmame (e.g., xmame.x11) and get past the legal screen.
  It looks like its the absence of a .cfg file that prevents xmame.xgl from
  running the game.  Thus far I have no idea why, though.
- Running with the Smotek aRts driver gives
  ** Warning ** unix_connect: can't connect to server 
  (unix:/tmp/mcop-chris/bagend-0b9c-3b1e3d10)
  arts_init error: can't connect to the aRts soundserver.
- Switching to fullscreen DGA mode and then quitting xmame screws up the
  keyboard.  Apparently this has something to do with xmame losing the ctrl or
  alt release codes; pressing ctrl or alt a few times should restore the
  keyboard to a sane state.
- Figure out how to handle mess_printf.  This may only exist in mess cvs right
  now.

Reported in 0.37b16.1:
- xmame.sdl doesn't work under BeOS; it just stalls.
- xmame.xgl will crash if you pause/unpause when running sfz3jr1 with
  -fullscreen and -bpp 16.  This could be specific to running the NVidia drivers
  with a GeForce 2 MX 64Mb card; I can't reproduce it with my Matrox G400.
- xmame.xgl will crash if you try to run it with a "NO GOOD DUMP KNOWN" rom.
  After crashing like this, subsequent attempts to run cps1/cps2 games will also
  crash it until X is restarted.  This might be a problem with NVidia's XFree86
  binary.
- Pausing/unpausing in fullscreen DGA mode will not allow the sound device to be
  reacquired.  The error displayed is "error: /dev/dsp: Device or resource
  busy."  After switching back to windowed mode, lsof shows /dev/dsp still being
  used by xmame, whether paused or unpaused.  Weird.  My sound card is an SB128
  PCI (es1370), so it's similar to Nick's Ensoniq AudioPCI (es1371); I'm not
  sure if this has anything to do with the problem.
- A couple people have reported that asterix and gradius3 run fine once and then
  never runs again, reporting that a couple of the ROMs are bad.  I haven't been
  able to reproduce this.  I wonder if it might have something to do with a bad
  .cfg file being written out.  One user (garph) reports it happening on NetBSD
  1.5 in plain X11 mode, so there's no exotic setup in common.  Update: Using
  -fakesound appears to make the problem go away.
- Hot Chase (hotchase) endlessly cycles through its initialization screens. This
  doesn't happen with mamew.  The solution is to use the asm 68k core instead of
  the C one.
- Update the docs to mention the problems using the nasm that comes bundled with
  Red Hat.  Encourage people to use a vanilla nasm 0.98 or higher.

Reported in 0.53.1:
- 8bpp mode is no longer supported by the core.  There may be a way for xmame to
  downgrade the 16bpp bitmap so that we can support 8bpp video modes, at least
  for games which use 256 or fewer colors.  For those which use more, we might
  be able to use a 332 RGB mode.
- One of the files uses a variable called "x11" which causes a compile error.
  Chris Stone thinks it's probably a macro in an X11 include file that's
  interfering.  If there isn't a straightforward way to fix it in xmame, I'll
  see if the core folks would be so kind as to rename the variable.
- Running xmame --version still tries to read the various .xmamerc files.  "Fix"
  this if it's simple to do.
- Change the joystick code so that if a numeral if something like -jdev
  /dev/input/js1 is specified, then that literal string is tried before
  appending 0, 1, 2, etc. to the device name.
- Running xmame.xgl with ssf2tu, sfa3, and spf2xj won't work if something
  changes the default.cfg (e.g., editing the controls).  They will work again
  after deleting default.cfg.

Reported in 0.54.1:
- See if nec765_dummy.c is necessary anymore in xmess.
- Figure out a way to force the use of DGA 1.0 on startup.
- Put back the -mono flag that existed around xmame-0.34.x.

Reported in 0.55.1:
- Look into adding XVideo extension support for scaling.  See MPlayer and
  gnuboy.
- Fix the tiny build for cvs mess.
- Williams games look messed up when using xmame.xgl -fullscreen.
- Add a note to the docs about Red Hat's nasm package screwing up builds.  Try
  to detect this case.

Reported in 0.55.2:
- Add a script for running under the problematic versions of KDE: 
    xmame.x11 -rid `xwininfo -root | grep xwininfo: | awk '{print $4}'` <game>

Reported in 0.56.1:
- Ask the core MAME folks to change the name of the PPC #define in v60.c to
  something that doesn't cause problems in LinuxPPC (e.g., _PPC).
- Try to figure out what -O2 with gcc 3.0.3 does to cause a segfault in narc.
- Come up with a test case for the gcc devs to duplicate the -O3 problem with
  z80.c
