|
From: Bryan W. <br...@vi...> - 2007-09-28 19:55:20
|
Hi folks,
I've tried several different binary distributions of valgrind under
CentOS 5, and now I've grabbed the latest valgrind source code and
compiled it myself. In all cases, I see the following behavior:
If I start up valgrind like:
valgrind --db-attach=yes myprog
and then type "y" to invoke gdb when prompted, gdb begins to
run but hangs before it ever gets to the (gdb) prompt. No
input is accepted (even CTRL-C and CTRL-Z). Looking at the
processes running at this stage, I see something like:
pid1 \_ valgrind --db-attach=yes myprog
pid2 \_ valgrind --db-attach=yes myprog
pid3 \_ /usr/bin/gdb -nw /proc/5882/fd/1014 5882
If I send pid2 a signal with "kill pid2", I then get a gdb
prompt.
The valgrind version is 3.2.3 and this is gdb version
gdb-6.5-16.el5, from CentOS 5. Invoking gdb by itself (outside
of valgrind) works fine.
Any ideas?
Thanks,
Bryan
|
|
From: Bryan W. <br...@vi...> - 2007-10-01 17:48:24
|
Bryan Wright <bryan <at> virginia.edu> writes: > If I start up valgrind like: > > valgrind --db-attach=yes myprog > > and then type "y" to invoke gdb when prompted, gdb begins to > run but hangs before it ever gets to the (gdb) prompt. No > input is accepted (even CTRL-C and CTRL-Z). Looking at the > processes running at this stage, I see something like: > > pid1 \_ valgrind --db-attach=yes myprog > pid2 \_ valgrind --db-attach=yes myprog > pid3 \_ /usr/bin/gdb -nw /proc/5882/fd/1014 5882 > > If I send pid2 a signal with "kill pid2", I then get a gdb > prompt. Hi again, One other datum I thought I'd share: Looking at pid3 (the gdb process) with strace, I see that it has called waitpid and is waiting for pid2 to receive a signal before continuing. This is consistent with the behavior I described above. Any idea why this is happening? Bryan |