@@ -1952,11 +1970,17 @@ fi
  # Note that WIN32APILIBS is set by GDB_AC_COMMON.
  WIN32LIBS="$WIN32LIBS $WIN32APILIBS"
  +# Object files to be used when building with support for all file formats.
+# This should not have elf or macho, as support for those formats depends
+# on BFD enabling them as well.
+all_binary_file_srcs="\$(dbx_SRCS) \$(mips_SRCS) \$(coff_SRCS) \$(xcoff_SRCS)"
+
+support_elf=no
  # Add ELF support to GDB, but only if BFD includes ELF support.
  GDB_AC_CHECK_BFD([for ELF support in BFD], gdb_cv_var_elf,
                   [bfd_get_elf_phdr_upper_bound (NULL)], elf-bfd.h)
  if test "$gdb_cv_var_elf" = yes; then
-  CONFIG_OBS="$CONFIG_OBS elfread.o stap-probe.o dtrace-probe.o \
+  CONFIG_OBS="$CONFIG_OBS stap-probe.o dtrace-probe.o \
          gcore-elf.o elf-none-tdep.o"

Does it make sense (or is it useful) to compile these other files (other
than elfread.o) if ELF support is not enabled?  Could we just list them
all in elf_SRCS so that they are not built if the user doesn't want ELF
support?  If you don't have an ELF reader.

I don't know. I took the path of least changes to implementing this, made worse by me not having a convenient way to test a non-ELF target.

I'll try to find a way to test if those are necessary and update the patch if they aren't

ah, turns out I could've just tried compiling with no elf support, oops. stap-probe and dtrace-probe can go

gcore-elf and elf-none-tdep are required to compile with --enable-targets=all, and since these aren't enough to consider elf supported, I think it's easier to just leave them here than figuring out which targets require them... I did notice that the arm-tdep.c does need them, but didn't go far into anything...

I'll submit a v7 with that update shortly

-- 
Cheers,
Guinevere Larsen
She/it