Re: PATCH for 64-bit MIPS ELF buglets

[Mark Mitchell <mark@codesourcery.com>]

>>>>> "Ian" == Ian Lance Taylor <ian@zembu.com> writes:

    Ian> What is going on here is that some people use elf32-mips when
    Ian> compiling ISA level III code.  This works fine, provided you
    Ian> always load the code at an address from 0 to 0x7fffffff or
    Ian> 0xffffffff80000000 to 0xffffffffffffffff.  The R_MIPS_64
    Ian> reloc in elf32-mips supports this usage: it is a 64 bit
    Ian> relocation formed by sign extending a 32 bit address.

OK.  I'll do what I suggested where I use different HOWTOs in 32-bit
vs. 64-bit mode for R_MIPS_64.

    Ian> There used to be code in mips_elf_relocate_section which
    Ian> handled this usage.  However, I see that you have removed it.

I'll see what it did, and restore that behvaior.

    Ian> I now see that you have also removed the support for
    Ian> R_MIPS16_26 while you shuffled code from
    Ian> mips_elf_relocate_section to mips_elf_calculate_relocation.

That wasn't really a shuffling; it was a total gutting. :-)

The loss of R_MIPS16_26/R_MIPS16_GPREL was an oversight on my part.  I
fully intended to restore them, but forgot.  However, I cannot find
any documentation on what these relocations do.  Can you point me at
anything, other than the previous code?  If you cannot, I will try to
decipher it, and make the right thing happen, but I would feel more
comfortable working from documentation.

    Ian> You must not assume that the MIPS ELF code is only used on
    Ian> native Unix systems.

Of course.  

You occasionally point out the obvious; I'm not clueless, just a bit
new to binutils.  When big rewrites happen, things tend to get broken
for a while; now that the rewriting is done, these things will get
fixed up.  Just the usual stabilization phase.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com