From mboxrd@z Thu Jan  1 00:00:00 1970
From: Andrew Cagney <ac131313@cygnus.com>
To: GDB Patches <gdb-patches@sourceware.cygnus.com>
Subject: Re: [rfc] Regcache revamp (vip)
Date: Sun, 18 Mar 2001 14:17:00 -0000
Message-id: <3AB3D655.E96868BD@cygnus.com>
References: <3AB2CE20.4AC725CF@cygnus.com>
X-SW-Source: 2001-03/msg00336.html

The change to read_register_bytes() included:

> !       if (reg_start >= in_start && reg_end <= in_end)
> !       /* register fits inside of in_buf. */
> !       memcpy (in_buf + (reg_start - in_start), reg_buf, reg_len);
> !       else if (reg_start >= in_start && reg_end > in_end)
> !       /* register crosses the end of in_buf. */
> !       memcpy (in_buf + (reg_start - in_start), reg_buf, in_end - reg_start);
> !       else if (reg_start < in_start && reg_end <= in_end)
> !       /* register crosses the start of in_buf. */
> !       memcpy (in_buf, reg_buf + (in_start - reg_start),
> !               reg_end - in_start);
> !       else
> !       internal_error (__FILE__, __LINE__, "read_register_bytes botch");

Hmm, this is missing one case - in_buf fits in register.  Lets instead
try:

      /* start = max (reg_start, in_start) */
      if (reg_start > in_start)
        start = reg_start;
      else
        start = in_start;

      /* end = min (reg_end, in_end) */
      if (reg_end < in_end)
        end = reg_end;
      else
        end = in_end;

      /* Transfer just the bytes common to both IN_BUF and REG_BUF */
      for (byte = start; byte < end; byte++)
        {
          in_buf[byte - in_start] = reg_buf[byte - reg_start];
        }

Andrew