Heh seems the assembler wasn't as tough as I thought it would be, I seem to have found a possible fix for the issue. Here's the relevant chunk again (code/io/timer.cpp lines 292-308):
#if defined(_MSC_VER)
// Timing in milliseconds.
_asm mov edx, temp_large.HighPart
_asm mov eax, temp_large.LowPart
//_asm shld edx, eax, 16 ; Keep 32+11 bits
//_asm shl eax, 16
// edx:eax = number of 1.19Mhz pulses elapsed.
_asm mov ebx, Timer_freq
// Make sure we won't divide overflow. Make time wrap at about 9 hours
sub_again:
_asm sub edx, ebx ; subtract until negative...
_asm jns sub_again ; ...to prevent divide overflow...
_asm add edx, ebx ; ...then add in to get correct value.
_asm div ebx
//eax = milliseconds elapsed...
_asm mov tmp, eax
tmp is an int, and the problem SEEMS to be that, on my system, at the end of this block eax is equal to a number around 33,000,000. This, when moved into tmp (which is a regular int) gets taken to be a negative number of about -9,000,000. And at that point everyone gets all sad. The solution I've found is to turn tmp into an unsigned int (as there's no way any of the system timer functions would EVER be returning a value less then zero according to MSDN specs), and also turn the "now" variable (in code/ship/afterburner.cpp line ) for the afterburner into an unsigned int as well. This worked.
Then I realized that all that NEEDS to be changed is the "now" variable, as the timer function will return a signed int that can be converted to unsigned without losing any data. Tried it, tested it, it works.
So to sum it up, changing the declaration of now (code/ship/afterburner.cpp line 242) from "int" to "unsigned int" seems to fix this problem on my machine. The scope of this change SEEMS to be restricted to the problem area alone, but I'd once again like to reiterate that I'm not very familiar with the project overall and it's theoretically possible that this could break something or many things.
But just maybe I fixed it
Dave
