Re: [PATCH v2 11/16] scripts/qapi: pull c_name from camel_to_upper to caller

Paolo Bonzini <[email protected]> writes:

> On 2/25/26 09:32, Markus Armbruster wrote:
>>      { 'enum': '__org.qemu_x-Enum', 'data': [ '__org.qemu_x-value' ] }
>> 
>> This tests a downstream extension prefix __RFQDN_.
>> 
>> As is, we generate
>> 
>>      typedef enum __org_qemu_x_Enum {
>>          ORG_QEMU_X_ENUM___ORG_QEMU_X_VALUE,
>>          ORG_QEMU_X_ENUM__MAX,
>>      } __org_qemu_x_Enum;
>> 
>> Without the .lstrip('_'), we'd generate
>> 
>>      typedef enum __org_qemu_x_Enum {
>>          __ORG_QEMU_X_ENUM___ORG_QEMU_X_VALUE,
>>          __ORG_QEMU_X_ENUM__MAX,
>>      } __org_qemu_x_Enum;
>> 
>> Meh.
>> 
>> Turns out this isn't on purpose.  Back then, the loop to map camel to
>> upper worked differently, and produced unwanted leading '_'.  For
>> instance, it mapped 'AbraCadabra' to '_ABRA_CADABRA'.  The .lstrip()
>> made the function produce 'ABRA_CADABRA'.  It also makes it eat
>> downstream extensions' leading '__'.  Oopsie.
>
> Which would be a reserved C identifier, wouldn't it?  So it seems it 
> still has its merit, after all (but it is limited to C).

Point.  There's still __org_qemu_x_Enum, though.