[torqueusers] Changing --with-debug to the default

David Beer dbeer at adaptivecomputing.com
Tue Jul 31 13:56:53 MDT 2012


Michael,

On Tue, Jul 31, 2012 at 12:52 PM, Michael Jennings <mej at lbl.gov> wrote:

> On Monday, 30 July 2012, at 17:26:10 (-0600),
> David Beer wrote:
>
> > Our support team has asked us repeatedly about making --with-debug
> > the default for TORQUE, making it so it can be disabled using
> > --without-debug, but obviously being on by default. Our support team
> > wants this so that when a site reports a core dump, the core has
> > information in it and the admin doesn't have to recompile and try to
> > reproduce in order to get enough information to debug the
> > crash. What are your thoughts on making this change?
>
> There are a few considerations here.
>
> 1 - Currently --with-debug eliminates all optimizations in $CFLAGS to
>     ensure that debugging symbols are accurate.  Some sites may object
>     to the potential performance penalty.  Many projects use -O2 -g
>     (or -O2 -g3) with reasonable success.  You may want to consider
>     splitting up the adding of -g3 (--with-debug) and the removal of
>     optimization (perhaps --without-optimization?).
>

Would you still think its worth it to separate these if I told you that the
default would be without optimization as well? The benefits of having
debugging symbols are compromised if optimizations are on, as you point out.


> 2 - RPM builds already (by default on most systems) generate separate
>     "debuginfo" packages which contain the debugging symbols for the
>     main package(s).  As mentioned above, while some symbols may be
>     marked as "optimized out," much of the debugging is still valid in
>     more recent gcc/gdb versions.
> 3 - I submitted a patch to Ken for a "--with debug" feature for the
>     spec file which would activate --with-debug AND allow the symbols
>     to remain inside the primary RPM packages instead of being moved
>     to the torque-debuginfo package.  It doesn't seem to have appeared
>     in SVN yet, so you may want to take another gander.
>
>
We will have to look into this patch. I don't believe I've seen it yet.


> That said, I'm all for making -g3 a default flag.  Just not sure about
> the rest.  ;-)
>
> Also, you may want to consider renaming --with-debug to --with-symbols
> or --with-gdb-symbols instead.  Having both --enable-debug and
> --with-debug (and having them do entirely disparate things) is rather
> confusing for new users.
>
>
Does anyone else have an opinion on this? I'm interested in making things
unambiguous.

David


> Michael
>
> --
> Michael Jennings <mej at lbl.gov>
> Senior HPC Systems Engineer
> High-Performance Computing Services
> Lawrence Berkeley National Laboratory
> Bldg 50B-3209E        W: 510-495-2687
> MS 050B-3209          F: 510-486-8615
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torqueusers
>



-- 
David Beer | Software Engineer
Adaptive Computing
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/torqueusers/attachments/20120731/875ccab9/attachment.html 


More information about the torqueusers mailing list