--------------------
Le Super Pendu (V1, words set #1) [Roberto Fresca, Grull Osgo, Jeff Hamelin]
Le Super Pendu (V1, words set #2) [Roberto Fresca, Grull Osgo, Jeff Hamelin]
This is not an actual user facing bug right now, because in the MAME UI the call to init_fs() is followed up by a call to setup_write(), which forces the image to commit without regard to the dirty bit. There is an argument that this is itself code smell; setup_write() perhaps should be set_output_format() and not arbitrarily perform a commit.
This will hopefully work around the very unhelpful uninitialised
variable warning that seems to be triggered by using Sol's get function
with std::optional<T> or sol::optional<T> on Linux.
* nes: Various minor fixes, more cleanups with bit functions.
bus/nes/benshieng.cpp: Streamline banking; no need to store variables in this device.
bus/nes/hes.cpp: Only support multicarts. Related singleton carts reassigned to relevant board types.
bus/nes_ctrl/hori.cpp: Amend some minutiae involving strobing and excessive controller reading.
hash/nes.xml: Removed a few more baddump flags.
* nes.xml: Replaced underdumped graphics ROM for Hanafuda Yuukyouden.
- Removed underdumped Soap Panic set.
- Replaced bad graphics ROM for Pokemon Silver hack.
- Trimmed overdumped PRG ROM to 4K.
- Removed nonexistent CHR ROM, replaced with emulation of on-board logic.
- Removed hack from NES slot code that directly set the CPU program counter.
- Corrected reset behavior.
This is problematic in several ways:
* Initialising things at construction that aren't needed until after
start slows down -romident, -validate, -listxml, etc. Slot cards can
be a particular drain on -listxml and -validate as they're
instantiated for every compatible slot. It's more pronounced for
array members, too.
* Splitting member initialisation between declaration in headers and
constructors in source files means you have to look at two places to
check for the initial value, and you always need to check the
constructor even if an initialiser is present in the header because
the constructor initaliser list takes precedence. (This isn't as much
of an issue for driver classes because the constructor is most often
inlined at declaration, so it isn't far from the member declarations.)
* Initialisers in headers for frequently-used devices increases the
frequency of recompiling dependent devices/drivers as they're exposed
to any changes in initialisers.
* Initialisers in frequently-used headers increase build times because
there's more for the compiler to parse/cache. (This affects
makedep.py as well for single-driver builds, but that's a single
pass.) It's not a lot individually, but it adds up given the size of
MAME, which keeps increasing. We've already had one contributor
banned from GitHub actions for resource usage, we don't want to waste
compiler time unnecessarily.
* Enabled GCC lifetime dead store elimination optimisation.
* emu/device.h: Don't pre-clear memory for drivers. Ivan Vangelista
fixed at least the majority of things that crashed outright, and
Robbbert initialised variables that coverity complained about. It's
unlikely anything will break due to this.
* sound/discrete.h: Explicitly initialise members of discrete "devices"
to zero. I don't see a way around doing this in headers due to the
macro soup used to build the constructors.
* sound/mos6581.cpp: Moved creation of the SID core to device_start and
explictly initialised members of the SID core structures. These
structures are in internal headers, so they won't cause downstream
recompiles.
-Lua engine: Made I/O port manager type_seq a bit more tolerant of
omitted arguments.
New working software list additions (nes.xml)
-----------------------------------
Chāozhí Gāo K Jīnkǎ 110 in 1 [Consolethinks]
8 in 1 (0801) [Consolethinks]
Golden 8 in 1 [Consolethinks]
