Ralf Baechle wrote: > On Fri, Apr 07, 2006 at 07:47:40PM +0200, Kevin D. Kissell wrote: > > > Arguably, whatever's used by binutils should be the tie-breaker. > > Googling around, I see that the EM_MIPS_RS3_LE value was > > added in the October 4, 1999 draft of the ELF spec, but inexplicably > > the alias with EM_MIPS_RS4_BE was left in place - perhaps they > > were supposed to be disambiguated by some 32-vs-64-bit flag > > somewhere. A random sampling of ELF documents on the web > > shows the vast majority calling out RS3_LE and not RS4_BE. > > No way to actually resolve this one; not even binutils oldtimer > Ian Lance Taylor can remember the reasons for the change anymore. FWIW, the general rule is to use EM_MIPS == 8 for MIPS with both endiannesses and ignore EM_MIPS_*. So removing the other defines from the linux include file seems to be the sensible thing to do. (Binutils can't do the same due to backward compatibility concerns.) Thiemo
