Hi,
We are facing an occasional core of our application which we think is
related to heap corruption. It is not reproducible (happens on some customers)
and takes weeks/months to occur.
Is there any way (in mdb or otherwise) to check memory heap of a user
process from core. I know libumem (with mdb) provides these. Most of our
customers run Solaris 8 and an OS upgrade is not an option for them.
One option would be to use a backport of libumem on Solaris 8, and use of
LD_PRELOAD to make the application use it instead of libc. The source code of
libumem is available. However, I am not sure about its dependencies on the
Solaris 9/10 architectural changes. Can anyone point out if its feasible, and
if yes then how it should be done.
Open to suggestions on other options for memory tracing, using other
libraries etc.
Thanks
Pari
This message posted from opensolaris.org
