Known issues :
- Tiled framebuffer layout is not yet implemented. This causes horizontal
offsets in some 2D 16bpp stuff (eg FMV in Tomb Raider 2, UI in Expendable)
- Fog is broken and currently disabled for Banshee
- Possibly at least 1 race in 2D blitter emulation
- A few things not yet emulated - overlay filters, some (rarely used) 2D
blitter functions
Patch from davefiddes. Original commit message :
This fixes the following warnings:
codegen_allocator.c:
- implict function declaration of rand() due to missing stdlib.h
codegen_backend_x86-64.c:
- implict function declaration of malloc() due to missing stdlib.h
- unused variables in codegen_backend_init()
- invalid typecast in use of host_x86_CALL()
codegen_backend_x86-64-uop.c:
- Remove unused variable declarations in codegen_CALL_FUNC_RESULT(),
codegen_LOAD_SEG(), codegen_OR_IMM(), codegen_XOR() and codegen_XOR_IMM()
- Move debug only variable declarations under the debug conditional define in
codegen_CALL_FUNC_RESULT(), codegen_LOAD_SEG(), codegen_MEM_LOAD_SINGLE(),
codegen_MEM_LOAD_DOUBLE(), codegen_MEM_STORE_SINGLE(),
codegen_MEM_STORE_DOUBLE()
codegen_reg.c:
- Change (int) typecasts to the correct (intptr_t) for small values stored
in a pointer. This fixes a series of 64-bit compilation warnings.
Tests:
- Build with no warnings on Fedora 32 with gcc 10.1
- Boot a Win98 machine with dynamic compilation enabled
Patch from davefiddes. Original commit message :
This fixes issues reported by gcc 10 with the size of various filename buffers
when concatenating strings into them.
Tests:
- Build cleanly with gcc 10
- Run a Pentium MMX Win98 machine and use custom GLSLP shader
Patch from davefiddes. Original commit message :
Remove several unused variable references which look like they were required for
earlier versions of the code.
Tests:
- Clean build on Fedora 32 with gcc 10.1
Patch from davefiddes. Original commit message :
The bank number is stored in a pointer value but is used exclusively as a
bank number integer reference. To do this safely it needs to be an intptr_t.
Patch from davefiddes. Original commit message :
GCC 10.1 reports this line as being logically always false so will never run.
On inspection it may be that this was intended to be:
if ((cyrix_addr & 0xf0) == 0xc0) return 0xff;
According to the IBM 6X86MX Microprocessor Databook
(http://datasheets.chipdb.org/IBM/x86/6x86MX/mx_full.pdf) this would correspond
to the Address Region 0 - 3 registers but omit Address Region 4 - 7 registers
which doesn't seem to make a lot of sense.
Delete the code as the default response from the function is to return 0xff
and this is what its behviour up until now has been.
