'const string&', which saves a string c'tor call each time they're
accessed.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2293 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
when it isn't needed. According to SDL this shouldn't be necessary, but
some combinations of video modes and audio hardware cause sound to stop
working if it's repeatedly opened and closed. So we have to keep it open.
Always attempt to use 2 channels (aka stereo) when opening the SDL audio
device. This is now required, since ROMs can request mono or stereo mode,
and the sound system can no longer be closed and re-opened with the
desired settings. So we need to always use 2 channels to accommodate
when it might be needed. This actually more closely emulates the 2600,
which does have two audio channels. The 'mono' vs 'stereo' ROM property
basically controls whether these channels are exposed (ie, does the
virtual console have two-channel audio hardware installed). In that sense,
these settings control how the channels are mixed.
Updated some settings to have more reasonable defaults.
Added acknowledgement for people that have donated hardware to help with
development.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2291 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
They were being saved correctly for all sticks, but only loaded for
the first one.
Removed some redundant settings related to TV effects stuff.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2289 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
ROMs with an odd # of scanlines will display the image in
black & white, just as on a real system.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2287 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
have actually been loaded during the program run. Otherwise,
it will erase previous mappings.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2286 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
saved per device, so that unplugging and replugging a joystick won't cause
the previous mappings to be lost.
All attributes of a joystick (ie, number of sticks, number of axes, buttons,
hats on each stick) are dynamic, meaning that hardcoded values are no
longer used, making the handling more robust.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2285 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
desktop on OSX and Windows, and the home directory in Linux.
The associated commandline has been changed to 'snapdir'.
Added 'GL VBO' toggle button to the Video Settings UI, and in
general cleaned up its interface.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2282 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
in this area.
Added commandline argument 'thumb.trapfatal', which causes fatal errors in
the Thumb ARM emulation to really be treated as fatal (ie, the emulation stops
and throws an exception). This is enabled by default (as it always should be).
When disabled, fatal errors simply log the error and continue with emulation.
This was added because the current HarmonyCart ARM code isn't always exactly
compatible with the emulation in Stella, and there is a lagtime from when
Harmony implements something to when the same functionality is added to Stella.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2279 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Specifically, the 3-voice audio routines read from ROM instead of RAM, and
the setting the ARM MAM register no longer causes the ARM emulation to crash.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2278 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
distinct GUI vs non-GUI (aka TIA) classes. This makes it easy to proceed
from this point on, since I don't have to worry about breaking one mode
when making changes to the other.
Preliminary support for TIA scanlines in OpenGL mode. It's disabled for
now, but is progressing nicely. This will be part of the new TV effects
in the next major release.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2273 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
In the process of converting the code to be OpenGL ES compliant, we shouldn't
hurt the performance on systems with more advanced versions.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2272 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
compliant. It actually went smoother than I thought it would.
This show allow Stella to run well on *any* hardware-accelerated
OpenGL machine, not just those with beefy video cards.
Cleaned up the code a little, removing more references to the old
OpenGL TV filtering (which will be replaced by Blargg NTSC
filtering soon enough). Also removed 'gl_texrect' option, which
specified to use the GL_TEXTURE_RECTANGLE_ARB extension (for
non power-of-two textures). It was never enabled by default,
didn't work with ATI cards, and wasn't accessible in the UI,
so I doubt anyone will miss it. Besides, it's not supported in
OpenGL ES 1.x.
Next is to look into VBO/PBO, to see if more optimizations can
be done.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2265 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
don't normally use the commandline (mostly Windows, but in some cases OSX as
well). The 'showinfo' commandline argument has been renamed 'loglevel', but
it has the same purpose. A new option 'logtoconsole' has been added, which
determines whether log output should also be directed to the commandline/
console (previously, it was always printed to the console). All these items
are now accessible from Options -> System Logs.
For anyone reading this (and that cares), now I can finally move on to the
OpenGL rewrite. The plan is that the new code will use OpenGL ES, which
is a subset of OpenGL 1.5. The main advantages are that you won't need
an advanced OpenGL card, and OpenGL ES is supported on most new 'smaller'
systems (iPhone, Android, etc), making ports much easier.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2264 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
BSPF_snprintf instead, which should eliminate any potential
buffer overflows. Also moved from static arrays to stringstreams
where appropriate to make things safer and more C++ like.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2263 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
of the C++ standard.
Fixed bug when using snprintf; we must use BSPF_snprintf, since the function
has different names in different OS's.
Optimized loading of PNG images in the RomInfoWidget so that memory allocations
aren't continuously being done. Basically, memory is allocated once and then
only re-allocated if a new image is larger than all previous ones. This can
increase memory use slightly, as the image data stays around between image loads,
and it remembers the largest image loaded. But if you images are all mostly the
same size, it should hurt memory usage too much. And it really helps on certain
systems where repeated (re)-allocations can cause problems.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2261 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Blargg NTSC filtering, but we need to revert to base functionality first.
Cleaned up the debugger API wrt converting values from integers to strings
(and vice-versa). There is now only one method for each of these functions,
rather than several ways to do so. In the process, convert from unsafe
sprintf functions into snprintf (eliminate potential buffer overflows).
The debugger 'print' command now indicates if any assigned label represents
a read-only location (R), a write-only location (W), or a read/write location
(R/W).
Fixed bug in DataGridWidgets (used in debugger ROM and CPU registers), where
scrolling the mouse-wheel would cause a segfault. Scrolling the mousewheel
now changes these items as expected.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2260 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
contents. The file was being accessed and read, but the commands it contained
weren't being executed! This bug appeared in April 2010, so it looks like
not many people are using this feature (or at least they're not reporting it).
Cleaned up a few compiler warnings in DebuggerParser class.
Bumped version # for beta release.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2254 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
for UNIX systems. Pathnames are always accessed internally as
full/absolute pathnames, and while they can be loaded and saved
with the '~' symbol (to indicate the users home directory), they
are still always mapped to absolute paths.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2252 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Fixed bug in handling analog axes with jitter; they were overriding events from
digital, hat and keyboard input.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2251 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
The debugger 'saverom' command now uses absolute filenames, and by default
will save data in the users home directory if a proper path isn't included
in the filename. This fixes a major bug where ROMs were being saved to
the current or application directory, which in some cases were invalid
locations.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2250 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
several cases (fixes problems when moving the mouse too fast or
when changing between video modes). Added extra signature to the
bankswitch autodetection for the 4A50 scheme.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2249 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
- it doesn't have a hotkey any more
- it is changed in the "Input Settings' UI, not in 'Video Settings'
- it only has meaning while in emulation mode
- it is enabled by default
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2245 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
That's why I'm not bumping the final version number until
just before the release ...
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@2238 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
stephena