This is the mail archive of the
ecos-discuss@sources.redhat.com
mailing list for the eCos project.
Re: [openrisc] Debugging application
- From: Scott Furman <sfurman at rosum dot com>
- To: openrisc at opencores dot org
- Cc: ECOS <ecos-discuss at sources dot redhat dot com>
- Date: Thu, 14 Aug 2003 07:24:52 -0700
- Subject: [ECOS] Re: [openrisc] Debugging application
- References: <NMEDLDELHPEFHOMFIJBHCELFCNAA.rcc@jennic.com>
Robert Cragie wrote:
Is it possible that there is an incompatibility with GCC 3.2.3 and GDB 5.0?
GDB 5.0 is rather old now. If so, do I need to try to use a later version of
GDB with OpenRISC additions?
Yes - there are apparently symbol name incompatibilities between gcc
3.2.3 and gdb 5.0. The situation is even worse if you try to debug C++
code since gdb 5.0 apparently does not support the compiler's revamped
mangling scheme.
For these reasons, I offered my port of OpenRISC gdb from version 5.0 to
5.3 on the openrisc list . (This is a straightforward port - with
minimal target-specific code changes.) I had no response to that
offer. If there's interest now and agreement from the OpenCores
maintainers, I can still check it in to OpenCores CVS. I've been using
gdb 5.3 for several months now and though it has its quirks, it seems to
work significantly better then gdb 5.0 on the OpenRISC target.
As far as seeing incorrect variable values, I find that the debugger is
occasionally mistaken about identifying the line of source code it is
about to execute. Stepping another instruction or two will make things
right. (This is a general problem I've noticed with gdb on many
platforms - not just OpenRISC.)
-Scott
--
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss