GDB partially works (for me)

Gene Smith gds at
Tue Aug 19 16:16:45 UTC 2008

Steven Grunza wrote:
> Given the large number of boards and types of processors supported by
> RTEMS, we need more info about your environment.  Details such as CPU
> and board would be helpful.

Bsp is derived from csb337 arm. My actual processor is arm946es. I have 
been looking at it a bit more and it appears that the timer interrupt 
does continue to occur after hitting a breakpoint but eventually 
_defaultExcHandler() from cpu.c is called. One time I saw 
_ISR_Nest_level at 5 and it just printed "System Locked" message. Other 
times it is zero and continuously prints that it is deleting the same 
thread. In both cases the aborted PC address is 0x201a0da0 which is 
_Watchdog_Sync_level in bss (not code!). So it goes into the weeds while 

Possibly interrupts need to be disabled while debugging/stepping?

Also, and I don't know if this is important, I see the message
in gdb "unsupport tag: DW_TAG_const_type" appear when I first load my 
gdbinit (when starting gdb or insight or reloading with "so gdbinit").

>> -----Original Message-----
>> From: rtems-users-bounces at 
>> [mailto:rtems-users-bounces at] On Behalf Of Gene Smith
>> Sent: Tuesday, August 19, 2008 10:33 AM
>> To: rtems-users at
>> Subject: GDB partially works (for me)
>> My system now runs OK (networking, serial, timers etc) and 
>> runs the test/demo programs OK and passes stress tests. If I 
>> set a breakpoint before leaving start.S I can hit the 
>> breakpoint and do the usual stuff. 
>> But if I try to continue after arriving at a breakpoint (such 
>> as at Init), nothing works (no timer, no networking no serial) 
>> and all I see happening is the idle task. Sometimes it will 
>> crash after (data abort
>> dump) this but usually just does idle loop.
>> I can look at all the registers that mask or turn off 
>> interrupts and they look OK. So I am not sure what causes it 
>> to completely shut down or go unstable after a b.p. is hit.
>> I do notice that sometimes I step into RTEMS code involving a 
>> watchdog (watchdogtickle). Not sure how this is used or if it 
>> might be shutting me down while at the breakpoint.
>> I get the same result with rtems4.8-gdb or insight. In both 
>> cases I am using a Macraigor MPDemon jtag debugger box.
>> -gene
>> _______________________________________________
>> rtems-users mailing list
>> rtems-users at
> _______________________________________________
> rtems-users mailing list
> rtems-users at

More information about the users mailing list