livie
/
phoenix-firestorm
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix LLVector4a::memcpyNonAliased16 so it does not overwrite memory past the end of destination buffer anymore.

meow-7.2.2
Nicky 2012-02-13 00:33:07 +01:00
parent d4579ba653
commit 7b03eb0083
1 changed files with 16 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
indra/llmath/llvector4a.cpp
View File

@ -27,6 +27,8 @@
#include "llmath.h"
#include "llquantize.h"
#include <string.h> // <FS:ND> for mempcy
extern const LLQuad F_ZERO_4A = { 0, 0, 0, 0 };
extern const LLQuad F_APPROXIMATELY_ZERO_4A = {
F_APPROXIMATELY_ZERO,
@ -80,12 +82,20 @@ extern const LLVector4a LL_V4A_EPSILON = reinterpret_cast<const LLVector4a&> ( F
}
}
while (dst < end)
{
copy4a(dst, src);
dst += 4;
src += 4;
}
// <FS:ND> There is no guarantee that the remaining about of bytes left is a number of 16. If that's not the case using copy4a will overwrite and trash memory behind the end of dst
// while (dst < end)
// {
// copy4a(dst, src);
// dst += 4;
// src += 4;
// }
bytes = (U8*)end-(U8*)dst;
if( bytes > 0 )
memcpy( dst, src, bytes );
// </FS:ND>
}
void LLVector4a::setRotated( const LLRotation& rot, const LLVector4a& vec )