On my machine (Mono) in Release config, this reduces the number of
string- and char-arrays allocated by an order of magnitude,
bringing the gamedb's total bytes allocated down by -10%.
AllStates() sorted the returned states, but this order was not used anywhere else, so I've removed it. GetStateClosestToFrame is solvable in O(n), so sorting is not required there either.
This is an O(n log(n)) -> O(n) improvement that is mostly relevant for the state history integrity checker
srl -> "official" DS rom extension (generally found in prototypes etc)
dsi -> used for DSi only games
ids -> used for iQue DS games
upstream handles all these rom extensions already
This fixes 3D Construction Kit and potentially other dsk games with more than 42 tracks on disc.
The +3 disc drive may only be able to read 41-ish tracks of standard sized data, but it appears many images have more tracks than this. 3D construction kit for example has 45 tracks in what passes for the TOC, but the last 5 of these are empty.
Make sure to scan entire directory record (some games seem to end up ending the search early with the old code)
Find "BOOT = cdrom:" with any variable amount of whitespace
This behavior "needed" for rcheevo hashing to work, since a few games (e.g. `Chinmoku no Kantai - The Silent Service (Japan)`, maybe others) have mode 2 form 2 sectors for the exe (for some reason)
Obviously we don't support IPF images right now, but with any luck we will eventually. So it makes sense to have something in place in the frontend to do core selection (when a gamedb hash is not found) based on the INFO block within the IPF file itself.
* Fixed RA hash for PlayStation games when BOOT line contains tabs where spaces are expected.
* Fixed RA hash and prevented infinite loop when PlayStation executable is in subdirectory.
there's some discrepancy between the 2 orders, because the GUI list relies on full names (of the groups) while in the code short names are used to add them to DB
build with `-p:MachineExtraCompilationFlag=BIZHAWKBUILD_DEBUG_RUMBLE`
there's nothing special about SDL2 I just can't be bothered finding the
other impls
This is most likely more correct. .net framework does not have a dedicated method for checking absolute paths, so this was most likely accidentally used instead.
-closes #4068
There's still two callsites that don't check the return value, but I'm not sure whether they can fail and what they should do if they do.
The Z80 should only be sampling the /WAIT pin during the 3rd cycle T of an IO REQ M Cycle.
Instruction timing tests verified on CPCHawk using WinAPE plustest.dsk.
Note: CPCHawk is the only core to currently use the Z80A /WAIT pin (FlagW), so other core exposure is 0.
* [CPCHawk] Start of new gate array and CRTC implementation
* [CPCHawk] Some cleanup
* [CPCHawk] More Cleanup
* [CPCHawk] More CRTC work
* [CPCHawk] More crtc work
* [CPCHawk] more CRTC work
* [CPCHawk] More CRTC work
* [CPCHawk] more crtc
* [CPCHawk] GateArray breaking changes WIP
* [CPCHawk] GA, CRTC and Screen changes
* [CPCHawk] CRTC and GateArray
* [CPCHawk] More stuff
* [CPCHawk] Progress
* [CPCHawk] Starting CRTC type abstraction
* [CPCHawk] Start CRTC status register implementation
* [CPCHAWK] Update colour palatte
* [CPCHawk] Trying to fix GA colour issues
* [CPCHawk] Actually use the Z80 /WAIT line
* [CPCHawk] Frame now running at req 19968 x6 INT
* [CPCHawk] Implement PAL16L8 as a separate device
* [CPCHawk] Border cropping SyncSetting
* [CPCHawk] Stuff
* [CPCHawk] More accurate CRT emulation
* [CPCHawk] Reset vertical timing var only when VSYNC ends in the middle of a scaline
* [CPCHawk] Allow IN port accesses to affect certain write-only devices
* [CPCHawk] Fixed high-impedence returns
* [CPCHawk] Readme update
* [CPCHawk] Update readme with Z80 timing test failures
* [CPCHawk] Update readme
Drawing a string is really having GDI+ do the drawing in software, uploading the result to a texture, then overlaying it to the render target. Since this texture is largely transparent, it needs blending. Of course, assuming the colors are all otherwise opaque, we only need to do this for the string texture draw
- setting multiple buttons to autofire actually works
- the config to respect lag in autofire is applied now
- can set meaningful axis hold / autofire
- removed last reference to "float autofire"
The documentation for these is lacking, as I've come to expect, but I'm
angry all the same.
For EmuHawk and DiscoHawk, I've pushed `$(Description)` (normally only
seen when publishing to NuGet) into `$(AssemblyTitle)`, which Windows
labels as "File description", and then explicitly set `$(Product)`.
Also `$(NeutralLanguage)` doesn't set a culture, it points any lookups
for `en-US` localisations to embedded resources, which I believe is the
default. This just makes it explicit.
broke with 8956657a34 I'm assuming. Notably this will now no longer copy the current settings over, but that might actually be more consistent with the behavior of the normal new option, so...
fixes#4056
`ControllerAdapter` ctor uses `(discs?.Length ?? 0) > 0`, but it gets
`discs` from earlier in core init; when it's later assigned to `_disks`,
empty lists are replaced with null, and other places in `NymaCore` use
`_disks != null`. The `default(int)` is to match pre-3af5b7a7f
behaviour.
in an ongoing effort to try and make the logic understandable
This should also restore default columns now when an existing tasmovie is currently loaded with column changes
and an existing bk2 is opened (which is then converted to a tasproj) (possible after 2ad7fc095c)
this allows loading the input roll settings stored in a tasproj even after it was initially `Load()`ed. This fixes those settings not getting applied when loading an existing, playing tasmovie into TAStudio.
If you want something like this done, just look for a .cue file next to the .bin file. Do not try to generate a temp .cue and pass that .cue along as the disc file to load, you will potentially run into a situation that temp .cue filename (mind you, this is random!) will act as the game name (due to not being present in db for whatever reason), affecting stuff like SaveRAM and such. The disc file "loaded" must not have some random junk in the filename.
ClearStickies clears buttons and axes, but the virtualpad tool wants to control whether only buttons are cleared.
also contains minor cleanups regarding related function calls.
YoshiRulz