> On Jul 21, 2020, at 5:17 AM, Pavel Labath <pa...@labath.sk> wrote:
>
> On 20/07/2020 23:38, Davide Italiano via lldb-commits wrote:
>> @@ -1601,8 +1601,18 @@ def single_step_only_steps_one_instruction(
>> # variable value
>> if re.match("s390x", arch):
>> expected_step_count = 2
>> + # ARM64 requires "4" instructions: 2 to compute the address (adrp,
>> add),
>> + # one to materialize the constant (mov) and the store
>> + if re.match("arm64", arch):
>> + expected_step_count = 4
>> +
>> self.assertEqual(step_count, expected_step_count)
>>
>> + # ARM64: Once addresses and constants are materialized, only one
>> + # instruction is needed.
>> + if re.match("arm64", arch):
>> + expected_step_count = 1
>> +
>> # Verify we hit the next state.
>> args["expected_g_c1"] = "0"
>> args["expected_g_c2"] = "0"
>>
>
> I have a feeling this was racing with aa73ee052f -- [lldb/test] Use
> inline assembly for instruction counting tests.
>
> There should be no need for magic step counts after that patch. (If
> there is, I'd like to know why.)
>
> pl
Interesting. I’ll try on my device and see whether we can revert this.
_______________________________________________
lldb-commits mailing list
lldb-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
