Re: [ECOS] diag_print problems by using multi-ice-gdb-server...

["benny" <benny5 at mindspring dot com>]

Hi,
I basically have the same problem when I enabled
CYGDBG_HAL_DEBUG_GDB_INCLUDE_STUBS.
My GDB is connected to the ICE but the debug serial port is still used by
the
GDB. I tried to disable mangling - didn't help.
The only way to eliminate this is not to include stubs. Then I'm able to
debug code
using multi-ICE (with multi-ICE basically you don't need any GDB support by
the
host because multi-ICE access memory directly through the JTAG).
Still I can't get the thread-awareness:
1. there seems to be bug in vector.s so the ice_thread_vector is not
initialized properly.
My GDB server don't see eCOS, it tells "Not a good vector or eCos not
running!"
The vector should be in some fixed location in memory known by the GDB
server, but I
don't know what this location should be.
After fixing bug in vector.s I got this vector at address 0x150 (0x50 to
0x14f is the
virtual vector table). I don't know if it right or wrong.
2. thread-awareness require CYGDBG_HAL_DEBUG_GDB_INCLUDE_STUBS, but this
creates problems
identical to what you have - on any attempt to use diag_print(), host hangs
waiting
for GDB to response (and there is no GDB on serial port).

If somebody can help ?

Regards,
Benny




>I made and installed the 'multi_ice_gdb_server' and it runs fine.
>An output of it means that comm-channels are provided but the
>first diag_print() call of my application fails. The insight (5.1) hangs
>there.
>If diagnostic outputs route to a serial port, the application runs fine to.
>Any ideas?
>
>Thanks in advance,
>Kai


-- 
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss