Gdb 6.8 just plain doesn't work on ARM targets. If a program is stopped in anything other than user-code (e.g. it's blocked in read() or accept() there's no backtrace. If the application is multi-threaded you see SIG32 messages. There are also reports of other odd behavior. Using 6.7.1 instead of 6.8 results in everything working. http://news.gmane.org/find-root.php?message_id=%3ch1omae%24i07%241%40ger.gmane.org%3e http://news.gmane.org/find-root.php?message_id=%3ch1pgtf%24oc2%241%40ger.gmane.org%3e http://news.gmane.org/find-root.php?message_id=%3ch1p5br%2411r%241%40ger.gmane.org%3e If you do a bit of searching you can find various other reports that 6.8 doesn't work on ARM and 6.7.1 works fine. My guess is that the problem underlying all of this is gdb's failure to figure out what library files are mapped. See the second link above for an example, but when you compare the output of "info files" to that of /proc/<pid>/maps, you'll see that library files (such as libc and libpthread) show up in the /proc/<pid>/maps but not in "info files". My guess is that causes the lack of backtrace, the SIG32, and all the other problems that people see on ARM using 6.8.
Do you still see these problems with recent releases?
Closing due to lack of response and age of the gdb in question.