Hi Rainer,
On 2026-03-06 10:57, Rainer Orth wrote:
Hi Torbjorn,
Below patch causes the following regressions for arm-none-eabi:
FAIL: outputs-22 exe savetmp namedb-2: outputs.ld1_args
FAIL: outputs-23 exe savetmp named2-2: outputs.ld1_args
FAIL: outputs-24 exe savetmp named2-3: outputs.ld1_args
FAIL: outputs-25 exe savetmp named2-4: outputs.ld1_args
FAIL: outputs-294 lto sing unnamed-3: a.ld1_args
FAIL: outputs-294 lto sing unnamed-3: a.ld_args
My gcc is configured with these flags:
https://github.com/STMicroelectronics/gnu-tools-for-stm32/blob/b8cdff71b827478c8510c70b3a15d2b50c759255/build-toolchain.sh#L291
https://github.com/STMicroelectronics/gnu-tools-for-stm32/blob/b8cdff71b827478c8510c70b3a15d2b50c759255/build-toolchain.sh#L417
Please let me know if you need any additional information.
it's usually best to file a PR and assign it to me.
In this case where a configure change is the culprit, it's best to
provide a diff of gcc/auto-host.h with and without the patch (and maybe
config.status if this gives anything more). This may save me building a
cross toolchain to check myself.
Besides, it would be good to have the actual failure/difference of one
of the failing tests.
That said, I've already had a look:
With my patch, auto-host.h again lost
@@ -1640 +1616 @@
-#define HAVE_GNU_AS 1
+#define HAVE_GNU_AS 0
@@ -1652 +1628 @@
-#define HAVE_GNU_LD 1
+#define HAVE_GNU_LD 0
I think I can simply revert this patch: setting gas_flag/gnu_ld_flag on
Solaris should already be handled by
Switch to HAVE_SOLARIS_AS
https://gcc.gnu.org/pipermail/gcc-patches/2026-March/709588.html
and
Switch to HAVE_SOLARIS_LD
https://gcc.gnu.org/pipermail/gcc-patches/2026-March/709589.html
In stage 1, I hope to get rid of the few remaining uses of
HAVE_GNU_AS/HAVE_GNU_LD (together with gas/gas_flag/gnu_ld/gnu_ld_flag)
as outlined in
https://gcc.gnu.org/pipermail/gcc-patches/2026-February/708353.html
https://gcc.gnu.org/pipermail/gcc-patches/2026-February/708355.html
to avoid this mess once and for all.
I'll test the reversal this weekend. Maybe you can file the PR
nonetheless?
Sorry for the late reply.
I see that you have already reverted the change, so I suppose you no longer
want a PR.
Anyway, I've just completed a test run with arm-none-eabi, and the regression
is no longer present as of r16-8253-geb50d28a9353e9.
Kind regards,
Torbjörn
Sorry for the breakage.
Rainer
