I believe the mangled line number section came from supplying a line
number of 0. I am now seeing this error when running opannotate (which
is causing serious thrashing on a 4core core 2 Intel box with nearly
4GB of RAM) reporting repeatedly:
unable to fork, error: Cannot allocate memory
An error occur during the execution of objdump:
I see no annotated assembly. Is there a way to stop opannotate from
swamping the box and surviving within my memory limits?
On 30 April 2010 12:22, Ian Rogers <rogers.email@...> wrote:
> I've created an option for the metacircular runtime platform
> (http://mrp.codehaus.org/) to generate opagent events so that I may
> debug its dynamically generated Java code (MRP is a project derived
> from Jikes RVM). On running opannotate I get messages of "BFD: Dwarf
> Error: mangled line number section.". I believe the reason is that the
> method or file names are in some way not conforming, but its hard for
> me to determine what exactly is broken. As well as dynamically
> generated code, pre-compiled code loaded from disk is also described
> to the opagent but I'm not seeing events in this code. If I add line
> number information to this loaded code (about 4MB worth) then my
> machine thrashes and ultimately, after several minutes of thrashing, I
> get no output from opannotate.
> I'm using oprofile 0.9.6 and would appreciate help debugging my problems.
> Many thanks,
> Ian Rogers