Zach Welch wrote: > On Sun, 2009-04-19 at 21:25 -0500, Dick Hollenbeck wrote: > >> [snip] >> Why are you changing this argument from the machine's natural size to >> one that is 32 bits wide? >> >> Dick >> > > Fair enough. Here is take four. > > This patch was produced as a band-aid to prevent bleeding if users try > to checkout and build the current head with those drivers enabled. > Without this patch, the new -Werror setting could cause them to get cut > on the sharp edges. > > There are a plethora of places in the OpenOCD tree where the types need > to be revisited systematically and cleaned up. This is not that patch. > > Cheers, > > Zach >
-int ft2232_read(u8* buf, int size, u32* bytes_read) +int ft2232_read(u8* buf, u32 size, u32* bytes_read) I would simply make that int ft2232_read(u8* buf, unsigned size, u32* bytes_read) or even int ft2232_read(u8* buf, unsigned size, unsigned* bytes_read) If we don't need the "32 bitness", it is best to let the compiler, and the programmer, use the machine's natural size. int and unsigned are simple. On a 64 bit machine, the "unsigned" and the "int" goes to 64 bits, but that is its natural size, and is generally the most efficient, easier than accessing 32 bits. Microsoft was insidious about injecting needless types into C and C++ code. When Java was designed, they made a point of slapping that idea in the face. u8* is cool, we need unsigned bytes. u32 is cool, but only when you have to have the variable be 32 bits wide, otherwise int or unsigned are better IMO when you don't care if it is 32 bits or 64 bits. The 64 bit days are here. Dick _______________________________________________ Openocd-development mailing list [email protected] https://lists.berlios.de/mailman/listinfo/openocd-development
