This is the mail archive of the binutils@sources.redhat.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: The ia64 linker problems.


>Err.. there must be a bug elsewhere.  There _must_ be a 
>relocation type.

This looks like an ABI problem to me.

The 0 reloc is coming from tc-ia64.c operand_match() for case IA64_OPND_TAG13
and IA64_OPND_TAG13b.  Looking at the ABI, I can't see any reloc for these
13-bit tag fields.  Perhaps the ABI assumption is that these must always be
resolved at assembly time, and hence do not require external relocs.  However,
bfd isn't going to resolve a symbol reference here unless I have an internal
reloc.  Should I try to create an internal reloc with no corresponding external
reloc?  Or should I report this as a bug to the ABI committee?

At the moment, we are effectively creating R_IA_64_NONE relocs, which "works",
but does not give the result that the programmer intended, because the tag
field is always set to zero.

Jim

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