Re: [PATCH] target/s390x: improve cpu compatibility check error message

[Claudio Fontana <cfontana@suse.de>]

On 3/14/24 20:44, Nina Schoetterl-Glausch wrote:
> On Thu, 2024-03-14 at 20:00 +0100, Claudio Fontana wrote:
>> some users were confused by this message showing under TCG:
>>
>> Selected CPU generation is too new. Maximum supported model
>> in the configuration: 'xyz'
>>
>> Try to clarify that the maximum can depend on the accel by
>> adding also the current accelerator to the message as such:
>>
>> Selected CPU generation is too new. Maximum supported model
>> in the accelerator 'tcg' configuration: 'xyz'
>>
>> Signed-off-by: Claudio Fontana <cfontana@suse.de>
>> ---
>>  target/s390x/cpu_models.c | 11 ++++++-----
>>  1 file changed, 6 insertions(+), 5 deletions(-)
>>
>> diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c
>> index 1a1c096122..0d6d8fc727 100644
>> --- a/target/s390x/cpu_models.c
>> +++ b/target/s390x/cpu_models.c
>> @@ -508,14 +508,14 @@ static void check_compatibility(const S390CPUModel *max_model,
>>  
>>      if (model->def->gen > max_model->def->gen) {
>>          error_setg(errp, "Selected CPU generation is too new. Maximum "
>> -                   "supported model in the configuration: \'%s\'",
>> -                   max_model->def->name);
>> +                   "supported model in the accelerator \'%s\' configuration: \'%s\'",
>> +                   current_accel_name(), max_model->def->name);
>>          return;
>>      } else if (model->def->gen == max_model->def->gen &&
>>                 model->def->ec_ga > max_model->def->ec_ga) {
>>          error_setg(errp, "Selected CPU GA level is too new. Maximum "
>> -                   "supported model in the configuration: \'%s\'",
>> -                   max_model->def->name);
>> +                   "supported model in the accelerator \'%s\' configuration: \'%s\'",
>> +                   current_accel_name(), max_model->def->name);
>>          return;
>>      }
>>  
>> @@ -537,7 +537,8 @@ static void check_compatibility(const S390CPUModel *max_model,
>>      error_setg(errp, " ");
>>      s390_feat_bitmap_to_ascii(missing, errp, error_prepend_missing_feat);
>>      error_prepend(errp, "Some features requested in the CPU model are not "
>> -                  "available in the configuration: ");
>> +                  "available in the accelerator \'%s\' configuration: ",
>> +                  current_accel_name());
>>  }
> 
> I wonder if these might not be confusing in other circumstances, e.g. when
> running with KVM and the Linux version lacks support for some feature.

Here you are referencing specifically the last hunk right? Ie the "Some features requested..." message.

> I think something along the lines of:
> 
> error_...(errp, "... supported by the current configuration ...", ...);
> error_append_hint(errp, "Consider using a different accelerator, a different QEMU version or, when using KVM, a different kernel");
> 
> would be better.

Interesting I'll try something along these lines.

> 
> I'm not sure about line breaks in error message, I like the better grepability
> of unbroken lines but the coding style guide doesn't mention anything.

better greppability in the log (as the error message in the log), or in the source code (or both)?
I am generally in favor of both, but there might be constraints on line length, although scripts/checkpatch.pl did not complain when I attempted this (I wonder if bug or feature).

docs/devel/style.rst on the code line length topic says:
"Lines should be 80 characters; try not to make them longer..."

and it does talk about exceptions. In the case of error message strings I think this could be one one of those exceptions.

In terms of logs, I did not find anything either, the most pertinent section should be "Error handling and reporting" in the same file,
but there is nothing about breaking up [or not] a single message in errors with newlines.

>>  
>>  S390CPUModel *get_max_cpu_model(Error **errp)
> 

Thanks,

Claudio