https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127256
--- Comment #10 from Drea Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Robin Dapp from comment #9)
> Valgrind says when compiling this test:
> I'd say this confirms that it's unrelated and a coincidence. I'll go ahead
> and post the patch and open a PR for this.
Yes I suspect this patch fixes the issue too:
```
diff --git a/gcc/config/aarch64/aarch64-json-tunings-parser.cc
b/gcc/config/aarch64/aarch64-json-tunings-parser.cc
index b2cfcfd3c02..05cfc9575dc 100644
--- a/gcc/config/aarch64/aarch64-json-tunings-parser.cc
+++ b/gcc/config/aarch64/aarch64-json-tunings-parser.cc
@@ -661,7 +661,7 @@ test_json_boolean ()
static const cpu_cost_table default_cost_table = {};
- tune_params params;
+ tune_params params{};
params.insn_extra_cost = &default_cost_table;
aarch64_load_tuning_params_from_json_string
```
Basically params was not zeroed before the call to
aarch64_load_tuning_params_from_json_string which means it was uninitialized
memory was being used. I might get some time tomorrow to test that.
