The problem is most noticeable when you are saving state, and the
save takes a relatively long time, short period of audio gets played
repeatedly, which is usually unpleasant.
I found out it's caused by sdl_sound::sdl_callback not fill the audio
buffer with silence when underflow occurs. According to
https://wiki.libsdl.org/SDL_AudioSpec, if there's nothing to play,
the callback should fill the buffer with silence.
I tested this change and the problem is gone.
Breakage was in 8338e0d7a4 (march 2015),
and yes, fullscreen didn't work correctly for me since then. Shows
how much I use it, I guess. FWTW my window manage is the venerable
fvwm2. CourierSud, if you happen to remember why you changed that (in
the middle of a lot of other changes), let me know, and we'll see how
to make it work for everybody.
* This turns the current state save feature to a menu; one can still press 0-9, but you can also browse a menu when loading and saving state
* Fixed some issues requested by Vas Crabb
* Updated state menu to support arbitrary character slots
* WIP, transitioned 'entry_char' to std::string
Still need to address Joystick and FR keyboard concerns
* Reimplemented state save support with joystick buttons
* Changed the state menu to be "code driven" rather than "character driven"
When the menu is displayed, it will look at the filenames, and translate them to a visual representation as per the user's locale
* Vas feedback
- Preliminary working DIO/DIO-II bus implementation and 98544 video card
- Hooked up i8042 I/O microcontroller comms to main CPU
- Support medium-res built-in video on 9000/332
this is the ctisland3 set that has been in MAME since forever. compared to the other ctisland sets it uses a different encryption and has a region code of 'D' in the header.
we haven't seen any 'D' region roms before (A is Japan, B is assumed to be US) so I've had to handcraft a bios rom for it (the only difference between the A/B bios roms is the single byte it checks in the header anyway)
D-CAT16 (Mega Drive handheld) [Team Europe]
this was dumped ages ago, doesn't boot as-is, maybe custom enhanced hw like the NES clones, maybe just custom banking.
I've discovered a scenario where reading to the end of file seems to trigger the fail bit, in addition to the eof bit. Because of this, I've changed the error message to display when we can't read from the stream, but the eof bit is _not_ set
