H8: Set H8/300H as the default architecture for H8 family
Checks
| Context |
Check |
Description |
| linaro-tcwg-bot/tcwg_binutils_build--master-arm |
success
|
Build passed
|
| linaro-tcwg-bot/tcwg_binutils_check--master-arm |
success
|
Test passed
|
| linaro-tcwg-bot/tcwg_binutils_build--master-aarch64 |
success
|
Build passed
|
| linaro-tcwg-bot/tcwg_binutils_check--master-aarch64 |
success
|
Test passed
|
Commit Message
Binutils, in particular ld and gas, use H8/300 as the default architecture.
On the other hand, support for H8/300 has been removed from gcc since
July 2020 (AFAIR) and H8/300H is the default there. This situation causes
that about 30 test cases fail in the following way:
[...]
Running /mnt/Works/binutils/ld/testsuite/ld-gc/gc.exp ...
h8300-elf-cc -B/mnt/Works/xcomp/build-binutils-h8300-linux/ld/tmpdir/ld/ -I/mnt/Works/binutils/ld/testsuite/ld-gc -g -O2 -fno-sanitize=all -c -ffunction-sections -fdata-sections -fno-sanitize=all -fno-lto -c /mnt/Works/binutils/ld/testsuite/ld-gc/gc.c -o tmpdir/gc.o
Executing on host: sh -c {h8300-elf-cc -B/mnt/Works/xcomp/build-binutils-h8300-linux/ld/tmpdir/ld/ -I/mnt/Works/binutils/ld/testsuite/ld-gc -g -O2 -fno-sanitize=all -c -ffunction-sections -fdata-sections -fno-sanitize=all -fno-lto -c /mnt/Works/binutils/ld/testsuite/ld-gc/gc.c -o tmpdir/gc.o 2>&1} /dev/null ld.tmp (timeout = 300)
spawn [open ...]
./ld-new -o tmpdir/gcexe -L/mnt/Works/binutils/ld/testsuite/ld-gc --gc-sections -e main --defsym __stack_chk_fail=0 tmpdir/gc.o
Executing on host: sh -c {./ld-new -o tmpdir/gcexe -L/mnt/Works/binutils/ld/testsuite/ld-gc --gc-sections -e main --defsym __stack_chk_fail=0 tmpdir/gc.o 2>&1} /dev/null ld.tmp (timeout = 300)
spawn [open ...]
/mnt/Works/xcomp/build-binutils-h8300-linux/ld/ld-new: h8300h architecture of input file `tmpdir/gc.o' is incompatible with h8300 output
/mnt/Works/xcomp/build-binutils-h8300-linux/ld/ld-new: h8300h architecture of input file `tmpdir/gc.o' is incompatible with h8300 output
FAIL: Check --gc-section
[...]
This patch sets H8/300H as the default architecture for H8 family.
bfd/ChangeLog:
* cpu-h8300.c: Set H8/300H as the default architecture
for H8 family.
gas/ChangeLog:
* config/tc-h8300.c: Set H8/300H as the default architecture
for H8 family.
ld/ChangeLog:
* configure.tgt: Set H8/300H as the default architecture
for H8 family.
Signed-off-by: Jan Dubiec <jdx@o2.pl>
---
bfd/cpu-h8300.c | 4 ++--
gas/config/tc-h8300.c | 2 +-
ld/configure.tgt | 4 ++--
3 files changed, 5 insertions(+), 5 deletions(-)
Comments
On 7/9/2026 3:04 PM, Jan Dubiec wrote:
> Binutils, in particular ld and gas, use H8/300 as the default architecture.
> On the other hand, support for H8/300 has been removed from gcc since
> July 2020 (AFAIR) and H8/300H is the default there. This situation causes
> that about 30 test cases fail in the following way:
I'm a lot less involved in binutils than I have been in the past, but
this makes sense to me. As Jan mentions, we dropped old H8/300 support
in GCC a while back and default to H8/300H.
Jan, do you have write permissions in the binutils repo?
Jeff
On 10.07.2026 02:53, Jeffrey Law via Binutils wrote:
> On 7/9/2026 3:04 PM, Jan Dubiec wrote:
>> Binutils, in particular ld and gas, use H8/300 as the default architecture.
>> On the other hand, support for H8/300 has been removed from gcc since
>> July 2020 (AFAIR) and H8/300H is the default there. This situation causes
>> that about 30 test cases fail in the following way:
> I'm a lot less involved in binutils than I have been in the past, but
> this makes sense to me. As Jan mentions, we dropped old H8/300 support
> in GCC a while back and default to H8/300H.
Imo NEW entries should be added for both gas/ and ld/.
Jan
On 10.07.2026 02:53, Jeffrey Law via Binutils wrote:
[...]> Jan, do you have write permissions in the binutils repo?
No, I don't.
/J.D.
On 10.07.2026 09:04, Jan Beulich wrote:
[...]> Imo NEW entries should be added for both gas/ and ld/.
Do you mean something like this:
gas/NEWS | 5 +++++
ld/NEWS | 5 +++++
2 files changed, 10 insertions(+)
diff --git a/gas/NEWS b/gas/NEWS
index 38928ab83ff..5f752169f86 100644
--- a/gas/NEWS
+++ b/gas/NEWS
@@ -24,6 +24,11 @@
* Add support for RISC-V vendor extensions:
SpacemiT: xsmtvdot v1.0, xsmtvdotii v1.0.
+* Change the default architecture for H8 family from H8/300 to H8/300H.
+ Support for H8/300 was removed from gcc in 2020, making H8/300H the
+ default architecture. The mismatch between the default architectures
+ used by gcc and gas/ld caused some ld test cases to fail.
+
Changes in 2.46:
* Add support for AMD Zen6 processor.
diff --git a/ld/NEWS b/ld/NEWS
index 2de87ae227c..32dc75f9ace 100644
--- a/ld/NEWS
+++ b/ld/NEWS
@@ -15,6 +15,11 @@ Changes in 2.47:
not only XCOFF. On by default and controlled with --no-link-mapless and
--link-mapless options, also for XCOFF.
+* Change the default architecture for H8 family from H8/300 to H8/300H.
+ Support for H8/300 was removed from gcc in 2020, making H8/300H the
+ default architecture. The mismatch between the default architectures
+ used by gcc and gas/ld caused some ld test cases to fail.
+
Changes in 2.46:
* Add --gnu-tls-tag/--no-gnu-tls-tag options to i386 ELF linker to add
/J.D.
On 10.07.2026 11:47, Jan Dubiec wrote:
> On 10.07.2026 09:04, Jan Beulich wrote:
> [...]> Imo NEW entries should be added for both gas/ and ld/.
> Do you mean something like this:
Almost; I'd ...
> gas/NEWS | 5 +++++
> ld/NEWS | 5 +++++
> 2 files changed, 10 insertions(+)
>
> diff --git a/gas/NEWS b/gas/NEWS
> index 38928ab83ff..5f752169f86 100644
> --- a/gas/NEWS
> +++ b/gas/NEWS
> @@ -24,6 +24,11 @@
> * Add support for RISC-V vendor extensions:
> SpacemiT: xsmtvdot v1.0, xsmtvdotii v1.0.
>
> +* Change the default architecture for H8 family from H8/300 to H8/300H.
> + Support for H8/300 was removed from gcc in 2020, making H8/300H the
> + default architecture. The mismatch between the default architectures
> + used by gcc and gas/ld caused some ld test cases to fail.
... suggest to omit the last sentence (also from ld/NEWS).
Jan
On 10.07.2026 12:04, Jan Beulich wrote:
> On 10.07.2026 11:47, Jan Dubiec wrote:
>> On 10.07.2026 09:04, Jan Beulich wrote:
>> [...]> Imo NEW entries should be added for both gas/ and ld/.
>> Do you mean something like this:
>
> Almost; I'd ...
OK then, I will post a new version of the patch in a few hours.
/J.D.
@@ -134,10 +134,10 @@ static const bfd_arch_info_type h8300s_info_struct =
N (32, 32, bfd_mach_h8300s, "h8300s", "h8300s", false, & h8300hn_info_struct);
static const bfd_arch_info_type h8300h_info_struct =
- N (32, 32, bfd_mach_h8300h, "h8300h", "h8300h", false, &h8300s_info_struct);
+ N (32, 32, bfd_mach_h8300h, "h8300h", "h8300h", true, &h8300s_info_struct);
const bfd_arch_info_type bfd_h8300_arch =
- N (16, 16, bfd_mach_h8300, "h8300", "h8300", true, &h8300h_info_struct);
+ N (16, 16, bfd_mach_h8300, "h8300", "h8300", false, &h8300h_info_struct);
/* Pad the given address to 32 bits, converting 16-bit and 24-bit
addresses into the values they would have had on a h8s target. */
@@ -52,7 +52,7 @@ int Smode;
int Nmode;
int SXmode;
-static int default_mach = bfd_mach_h8300;
+static int default_mach = bfd_mach_h8300h;
#define PSIZE (Hmode && !Nmode ? L_32 : L_16)
@@ -313,8 +313,8 @@ ft32-*-*) targ_emul=elf32ft32
targ_extra_ofiles=ldelfgen.o
;;
h8300-*-elf* | h8300-*-rtems*)
- targ_emul=h8300elf;
- targ_extra_emuls="h8300helf h8300self h8300hnelf h8300snelf h8300sxelf h8300sxnelf"
+ targ_emul=h8300helf;
+ targ_extra_emuls="h8300self h8300hnelf h8300snelf h8300sxelf h8300sxnelf h8300elf"
;;
h8300-*-linux*)
targ_emul=h8300elf_linux;