Re: [PATCH 1/3] Move the traceframe_available_memory code from memory_xfer_partial_1 down to the targets

[Pedro Alves <palves@redhat.com>]

On 03/11/2014 12:42 PM, Yao Qi wrote:

> 	(remote_read_bytes): Moved from memory_xfer_partial_1.

Say:

 	(remote_read_bytes): New, factored out from memory_xfer_partial_1.

> @@ -6847,6 +6928,63 @@ remote_read_bytes (CORE_ADDR memaddr, gdb_byte *myaddr, ULONGEST len,
>    if (len == 0)
>      return 0;
>  
> +  if (get_traceframe_number () != -1)
> +    {
> +      VEC(mem_range_s) *available;
> +
> +      /* If we fail to get the set of available memory, then the
> +	 target does not support querying traceframe info, and so we
> +	 attempt reading from the traceframe anyway (assuming the
> +	 target implements the old QTro packet then).  */
> +      if (traceframe_available_memory (&available, memaddr, len))
> +	{
> +	  struct cleanup *old_chain;
> +
> +	  old_chain = make_cleanup (VEC_cleanup(mem_range_s), &available);
> +
> +	  if (VEC_empty (mem_range_s, available)
> +	      || VEC_index (mem_range_s, available, 0)->start != memaddr)
> +	    {
> +	      enum target_xfer_status res;
> +
> +	      /* Don't read into the traceframe's available
> +		 memory.  */
> +	      if (!VEC_empty (mem_range_s, available))
> +		{
> +		  LONGEST oldlen = len;
> +
> +		  len = VEC_index (mem_range_s, available, 0)->start - memaddr;
> +		  gdb_assert (len <= oldlen);
> +		}
> +
> +	      do_cleanups (old_chain);
> +
> +	      /* This goes through the topmost target again.  */
> +	      res = memory_xfer_live_readonly_partial (&remote_ops,

This "&remote_ops" here will be wrong in case we're connected
with extended-remote.  Looks like remote_read_bytes needs to be
gain an ops parameter.

Otherwise looks good.

Thanks,
-- 
Pedro Alves