Re: [PATCH v4 2/2] ld: Add minimal pdb generation

[Mark Harmstone <mark@harmstone.com>]

On 12/10/22 01:28, Alan Modra wrote:
> On Tue, Oct 11, 2022 at 06:53:32PM +0100, Mark Harmstone wrote:
>> +/* Calculate the hash of a given string.  */
>> +static uint32_t
>> +calc_hash (const char *data, size_t len)
>> +{
>> +  uint32_t hash = 0;
>> +
>> +  while (len >= 4)
>> +    {
>> +      hash ^= *(uint32_t *) data;
>> +      data += 4;
>> +      len -= 4;
>> +    }
>> +
>> +  if (len >= 2)
>> +    {
>> +      hash ^= *(uint16_t *) data;
>> +      data += 2;
>> +      len -= 2;
>> +    }
>> +
>> +  if (len != 0)
>> +    hash ^= *data;
>> +
>> +  hash |= 0x20202020;
>> +  hash ^= (hash >> 11);
>> +
>> +  return hash ^ (hash >> 16);
>> +}
> I think the above code will calculate different hash values on
> big-endian machines to little-endian.  Also, unless "data" is aligned
> as for uint32_t you run the risk of alignment traps on machines with
> strict alignment requirements.
>
Okay, thanks Alan. Can we get the first patch accepted while I investigate this?

Mark