Steven Johnson wrote:
Registers in the remote protocol are Hex Encoded. Hex encoded values
can have (as far as I can tell, valid values of
'0'-'9','a'-'f','A'-'F' and ('x' for registers). the problem is that
register packets that have an upper case 'A'-'F' in the first
location are junked as being bad packets, when their is nothing
wrong. And then GDB ends up in an infinite comms loop, trying to
recover.
The attached patch allows Hex Encoded values to include upper case
letters (in the case of fetching registers) without causing the
packet handling to fail.
I wasnt sure if 'X' should also be allowable, seems like it should,
but i dont know for sure, so havent changed it.
Steven Johnson
What Andrew said, but OTOH, this is certainly not an intrusive change.
Steven, what remote target are you seeing this with?
[Someone will say "but then we'll have to make sure that we consistantly
accept upper case throughout, and I agree that's a little more intrusive]
------------------------------------------------------------------------
diff -p -u -r clean-src/insight-6.1/gdb/frame.c
mod-src/insight-6.1/gdb/frame.c
diff -p -u -r clean-src/insight-6.1/gdb/remote.c
mod-src/insight-6.1/gdb/remote.c
--- clean-src/insight-6.1/gdb/remote.c 2004-02-26
06:41:00.000000000 +1000
+++ mod-src/insight-6.1/gdb/remote.c 2004-06-15 16:36:53.000000000
+1000
@@ -3278,6 +3278,7 @@ remote_fetch_registers (int regnum)
and try to fetch another packet to read. */
while ((buf[0] < '0' || buf[0] > '9')
&& (buf[0] < 'a' || buf[0] > 'f')
+ && (buf[0] < 'A' || buf[0] > 'F')
&& buf[0] != 'x') /* New: unavailable register value */
{
if (remote_debug)