This adds support for marking RL78 binaries as either supporting 32-bit

or 64-bit doubles.  It also makes the linker complain if the user attempts
to link together binaries with different sized doubles.

	* elf32-rl78.c (rl78_elf_merge_private_bfd_data): Complain if
	64-bit doubles objects mix with 32-bit doubles objects.
	(rl78_elf_print_private_bfd_data): Describe 64-bit doubles flag.

	* readelf.c (get_machine_flags): Handle RL78 64-bit doubles flag.

	* config/tc-rl78.c (enum options): Add OPTION_32BIT_DOUBLES
	and OPTION_64BIT_DOUBLES.
	(md_longopts): Add -m32bit-doubles and -m64bit-doubles.
	(md_parse_option): Parse -m32bit-doubles and -m64bit-doubles.
	(md_show_usage): Show all of the RL78 options.
	(rl78_float_cons): New static functions.
	(md_pseudo_table): Update handler for "double".

1 files changed, 12 insertions, 2 deletions

diff --git a/gas/doc/c-rl78.texi b/gas/doc/c-rl78.texi
index dce833bc75..a7145482d9 100644
--- a/gas/doc/c-rl78.texi
+++ b/gas/doc/c-rl78.texi
@@ -32,6 +32,14 @@ Enable support for link-time relaxation.
 Mark the generated binary as targeting the G10 variant of the RL78
 architecture.
 
+@item m32bit-doubles
+Mark the generated binary as one that uses 32-bits to hold the
+@code{double} floating point type.  This is the default.
+
+@item m64bit-doubles
+Mark the generated binary as one that uses 64-bits to hold the
+@code{double} floating point type.
+
 @end table
 
 @node RL78-Modifiers
@@ -85,8 +93,10 @@ In addition to the common directives, the RL78 adds these:
 @table @code
 
 @item .double
-Output a constant in ``double'' format, which is a 32-bit floating
-point value on RL78.
+Output a constant in ``double'' format, which is either a 32-bit
+or a 64-bit floating point value, depending upon the setting of the
+@option{-m32bit-doubles}|@option{-m64bit-doubles} command line
+option.
 
 @item .bss
 Select the BSS section.