Re: [PATCH v2 14/19] read/write_pieced_value: Improve logic for buffer allocation

[Simon Marchi <simon dot marchi at polymtl dot ca>]

On 2017-05-09 19:46, Andreas Arnez wrote:
> @@ -2045,21 +2022,44 @@ write_pieced_value (struct value *to, struct
> value *from)
>  	  }
>  	  break;
>  	case DWARF_VALUE_MEMORY:
> -	  if (need_bitwise)
> -	    {
> -	      /* Only the first and last bytes can possibly have any
> -		 bits reused.  */
> -	      read_memory (p->v.mem.addr + dest_offset, buffer.data (), 1);
> -	      read_memory (p->v.mem.addr + dest_offset + this_size - 1,
> -			   &buffer[this_size - 1], 1);
> -	      copy_bitwise (buffer.data (), dest_offset_bits % 8,
> -			    contents, source_offset_bits,
> -			    this_size_bits,
> -			    bits_big_endian);
> -	    }
> +	  {
> +	    CORE_ADDR start_addr = p->v.mem.addr + dest_offset_bits / 8;
>
> -	  write_memory (p->v.mem.addr + dest_offset,
> -			source_buffer, this_size);
> +	    if (dest_offset_bits % 8 == 0 && this_size_bits % 8 == 0
> +		&& source_offset_bits % 8 == 0)
> +	      {
> +		/* Everything is byte-aligned; no buffer needed.  */
> +		write_memory (start_addr,
> +			      contents + source_offset_bits / 8,
> +			      this_size_bits / 8);
> +		break;
> +	      }
> +
> +	    this_size = bits_to_bytes (dest_offset_bits, this_size_bits);
> +	    buffer.reserve (this_size);

Reading Pedro's patch about byte_vector:

https://sourceware.org/ml/gdb-patches/2017-06/msg00339.html

reminded me of this, when I read the patch I didn't think it was a 
problem.  I think this should use resize instead of reserve.

Simon