Re: [PATCH] gdb, amd-dbgapi-target: use PRIu64 and PRIu32

[Pedro Alves <pedro@palves.net>]

On 2026-07-02 21:15, Tom Tromey wrote:
>>>>>> Lancelot SIX <Lancelot.Six@amd.com> writes:
> 
>> Hi,
>> Doesn't GDB have an implicit habit/policy not to use the PRI??? macros?
> 
>> I do not mind them in amd-dbgapi-target.c, but want to make sure
>> global maintainers are fine with this.
> 
>> I guess the alternative is to use pulongest + %s instead of PRIu64.
> 
>> i.e. something like this:
> 
>>     str += (agent_id != AMD_DBGAPI_AGENT_NONE
>>         ? string_printf (" %s", pulongest (agent_id.handle))
>>         : "?");
> 
> I don't know if it's a policy exactly, but personally I prefer pulongest
> et al, since I find the PRI* macros pretty ugly and unreadable.

Agreed.  And it's also a portability hazard, like other printf formats, as you have
to match the printf format to the type of the variable.  For typedefs, you have
to either hardcode the underlying type (as here) when it's supposed to be an implementation
detail, or add new PRI-like macros for the typedefs.  The pulongest (and paddress, etc.)
pattern mostly eliminates that, as the printf format is always "%s", at a small runtime cost.

(std::cout << ... << ... tends to be unreadable too IMHO, but we don't use that, thankfully).

{fmt} would be better, but that's either a new external dependency, or 
bump to c++20 for std::format, or c++23 for widely available std::print.

Pedro Alves