This is the mail archive of the binutils@sourceware.cygnus.com mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: MIPS ELF embedded-pic support.


   Date: Sat, 30 Oct 1999 17:31:06 +1000
   From: Geoff Keating <geoffk@ozemail.com.au>

   > I'm not comfortable with these changes.  You appear to be changing the
   > behaviour for STO_MIPS16, S_IS_WEAK, and linkonce symbols, and perhaps
   > others.  Is that wise?  Have you tested those cases thoroughly?

   I've run the g++ test suite, which will check the weak and linkonce
   cases; nothing terrible happened.

That is fairly minimal testing, unfortunately.  It does not test ld
-r, which is one of the more common cases to break.

   If someone could write a test case for STO_MIPS16 (and it didn't
   work), I could fix it.  I expect, though, that you'd fix it somewhere
   else than in this code; it's easier to have 'value' have some
   consistent meaning after this code (although it doesn't before it),
   and then use that in subsequent processing.

It's hard to fix it elsewhere.  The BFD actions tend to be constrained
by the linker.  There isn't anywhere else to change.

Relocation processing is very broken.  No question.  But fixing it
based on trying to acheive consistency in md_apply_fix is not going to
work.  That's where it is broken.  It needs to be fixed by setting up
different infrastructure in BFD and gas.  See the discussion in
bfd/doc/bfdint.texi.

Ian

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]