This is the mail archive of the
gdb@sourceware.org
mailing list for the GDB project.
stack trace problem for target arm
- From: "Johan Thorb" <johan dot thorb at gmail dot com>
- To: gdb at sourceware dot org
- Date: Wed, 17 Jan 2007 16:25:20 +0100
- Subject: stack trace problem for target arm
Hi All,
Anyone knowing how to fix a better stack trace for the following setup?
gdb-host: winXP
target: arm9
compiler for debugged binary: non-gcc (IAR) mixed arm/thumb-code
debugging with gdb-6.6-arm-eabi (also tried gdb-6.6-arm-thumb-elf)
Scenario:
Starting at main stack trace is ok,
- Stepping into the first funktion (both step & next results in a
step into) results in a corrupt stack trace. At this point the stack
trace shows 3 frames - the one executing on the top, but then 2
identical but wrong frames. These 2 frames shows the current address
and a not correct symbol. The (main-)function from where the call was
made is not shown.
- Stopping at a breakpoint at the end of the function -> the stack
trace shows 2 entries both with the current function symbol, but
different line numbers (the end source code row and a source code row
some lines above).
- Stepping one step (or next) back to main -> stack trace shows only
the main-frame (ok!).
This scenario repeates for other function calls futher ahead in main().
The code debuggd, mentioned above, is mainly thumb2 code.
problems: 1) backtrace indicate that stacktrace gets corrupt and
shows identical frames
2) 'next' steps info function (not always but often)
instead of stepping over (due to problem 1!?)
Would it be possible to fix gdb to handle this situation or is the
problem not solvable in gdb due to missing or incorrect information
in the elf-file?
Thanks in advance.
Best Regards,
/Johan