NetBSD/gnu/lib/bfd/_mips.mk

32 lines
1.0 KiB
Makefile

# $NetBSD: _mips.mk,v 1.7 1997/10/19 10:44:54 jonathan Exp $
BFD_MACHINES = cpu-mips.c
BFD_BACKENDS = elf32-mips.c elf32.c elf.c elflink.c ecofflink.c \
coff-mips.c ecoff.c
ARCH_DEFS = -DARCH_mips \
-DSELECT_ARCHITECTURES='&bfd_mips_arch'
TDEFAULTS = \
-DDEFAULT_VECTOR=bfd_elf32_littlemips_vec \
-DSELECT_VECS=' &bfd_elf32_littlemips_vec, &bfd_elf32_bigmips_vec, &ecoff_little_vec, &ecoff_big_vec ' \
-DHAVE_bfd_elf32_littlemips_vec \
-DHAVE_bfd_elf32_bigmips_vec \
-DHAVE_ecoff_little_vec \
-DHAVE_ecoff_big_vec \
-DNETBSD_CORE
# XXX cannot support 64-bit BFD targets with gdb 4.16.
# They assume that BFD_ARCH_SIZE is 64, but that causes bfd_vma_addr
# to be a 64-bit int. GDB uses bfd_vma_addr for CORE_ADDR, but also
# casts CORE_ADDRS to ints, which loses on 32-bit mips hosts.
# TDEFAULTS += \
# -DHAVE_bfd_elf64_bigmips_vec \
# -DHAVE_bfd_elf64_littlemips_vec \
# _DSELECT VECS = &bfd_elf64_bigmips_vec, &bfd_elf64_littlemips_vec,
#
# BFD_BACKENDS += \
# elf64-mips.c elf64.c
OPCODE_MACHINES = mips-dis.c mips-opc.c mips16-opc.c