This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: malloc probes for Systemtap
- From: Roland McGrath <roland at hack dot frob dot com>
- To: Alexandre Oliva <aoliva at redhat dot com>
- Cc: Andreas Jaeger <aj at suse dot com>, libc-alpha at sourceware dot org
- Date: Tue, 2 Oct 2012 15:37:13 -0700 (PDT)
- Subject: Re: malloc probes for Systemtap
- References: <ork3w94at9.fsf@livre.localdomain><20407326.zUUZoqzhOz@byrd><ormx05h0ez.fsf@livre.localdomain>
Sorry, I'm still behind on a lot of backlog. At a quick glance, there
is nothing wrong with the use of the LIBC_PROBE macro, which is the
thing here on which I'm the particular expert. I'm not at all an
expert on malloc implementation details, which is what goes to what
the useful set of probes to have is.
I commend your attention to documentation. But we have never before
described sdt probes in the manual, and have not yet established
conventions for how we describe them. What we have in the way of
documentation for the existing probes is just
nptl/DESIGN-systemtap-probes.txt and elf/rtld-debugger-interface.txt.
(There are also setjmp/longjmp probes, which aren't documented
anywhere that I recall.) It's not appropriate to put them into the
manual proper until we have both decided on the formatting conventions
for documenting probes in Texinfo and decided on the ABI stability
intentions for probes. (If they are not a stable ABI, then it may not
be appropriate to describe them in the manual at all.)
So, you need to get someone who is expert on malloc details to review
your choice of probe locations, names, and parameters. And, we need
to have a general discussion here about documentation and ABI
stability for probes. (There was some past discussion about this when
the earlier probes went in, but it only resolved to the point that we
achieved consensus that we didn't really need to resolve it right then.)
Thanks,
Roland