ShuriZma
/
desmume
mirror of https://github.com/TASEmulators/desmume.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

fix new bugs in save code which broke some games in sneaky ways. fixes apollo justice, although reportedly it was broken for other reasons in previous releases. also might have fixed a regression in tomb raider underworld

This commit is contained in:
zeromus 2013-12-01 06:52:14 +00:00
parent 0191b680c4
commit 912355488e
1 changed files with 22 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
desmume/src/mc.cpp
View File

@ -714,18 +714,25 @@ u8 BackupDevice::data_command(u8 val, u8 PROCNUM)
addr_counter++;
val = 0xFF;
if (addr_counter >= addr_size)
{
//why does tomb raider underworld access 0x180 and go clear through to 0x280?
//should this wrap around at 0 or at 0x100?
if(addr_size == 1) addr &= 0x1FF;
//if we just finished building the address, seek the FP
if (addr_counter == addr_size)
fpMC->fseek(addr, SEEK_SET);
}
}
else
{
ensure(addr);
//why does tomb raider underworld access 0x180 and go clear through to 0x280?
//should this wrap around at 0 or at 0x100?
//TODO - dont other backup memory types have similar wraparound issues?
if(addr_size == 1)
{
addr &= 0x1FF;
//since we changed the address unexpectedly (non-sequentially), be sure to seek the FP
fpMC->fseek(addr, SEEK_SET);
}
//if we're writing a byte at address 0, we need to ensure that we have at least 1 byte in the file (IOW, ensure to size addr+1 == 0+1)
ensure(addr+1);
if(com == BM_CMD_READLOW)
{
val = read();
@ -911,7 +918,13 @@ void BackupDevice::ensure(u32 addr, u8 val, EMUFILE_FILE *fpOut)
fp->fflush();
fp->fseek(addr, SEEK_SET);
//this is a HORRIBLE IDEA.
//leave the FP positioned to write the final byte
//this is a HACK to make the basic read/write byte operation work when it calls ensure().
//IDEALLY, no assumptions about the file pointer can be made.
//but someone (actually, not really) so very carefully profiled the save IO code and discovered that not fseeking for every byte read/write was a great optimization.
//so, now all this code is depending/assuming on the FP being kept in a precise position, and I dont think its smart to change the main user of this assumption to paper over this bug by making it fseek before read/write, while leaving other unknown assuming clients intact
fpMC->fseek(addr-1, SEEK_SET);
#endif
}