Re: [PATCH 4/6] backends/: should use size_t n_regs_mapping for *_sample_sp_pc

[Serhei Makarov]

On Sun, Apr 5, 2026, at 8:24 PM, Aaron Merey wrote:
>> @@ -84,8 +84,8 @@ generic_sample_sp_pc (const Dwarf_Word *regs, uint32_t 
>> n_regs,
>>    /* TODO: Register locations could be cached and rechecked on a
>>       fastpath without needing to loop, though the overhead reduction
>>       is minimal.  */
>> -  int j, need_sp = (sp != NULL), need_pc = (pc != NULL);
>> -  for (j = 0; (need_sp || need_pc) && n_regs_mapping > (uint32_t)j; j++)
>> +  int need_sp = (sp != NULL), need_pc = (pc != NULL);
>> +  for (size_t j = 0; (need_sp || need_pc) && n_regs_mapping > j; j++)
>>      {
>>        if (n_regs < (uint32_t)j) break;
>
> This cast will truncate j when size_t is uint64_t, although having
> more than 2^32 registers is unlikely.
Agreed, don't see a better way to do this.
The comparison will be fixed to use <= in the revised patch (cf the discussion 
on patch1),
so even in a deliberately-pathological case we would have j reach UINT32_MAX
and trigger this correctly.