oprofile 0.0.8 has been released

[John L. <le...@mo...>]

OProfile is a powerful and flexible profiler for Linux x86 systems.

oprofile 0.0.8 has been released, available as usual at http://sf.net/projects/oprofile/

Users are encouraged to upgrade, as this release fixes possible segfaults,
and inaccurate profiling data.
 
Feature improvements
--------------------

A new option has been added to allow samples inside libraries to be separated-out
into per-binary sample files. This means you can now differentiate between
samples in the C library caused by e.g. X, and samples caused directly by
your application, for example. This option has an attached additional overhead,
but tests have shown it is reasonable and unlikely to affect any real systems
too noticably. 
 
The daemon will now cleanly shutdown on an "op_stop", writing all pending profiling
data.
 
The utility "op_time" has been added for broad overviews of profiling data.
 
Documentation has been updated and improved.
 
Old sample files from old binary images are now backed up instead of deleted.
 
oprofpp can exclude a set of symbols from its list.
 
Various performance improvements have been made.
 
Bugfixes
--------

A bug that lead to fewer interrupts being generated when using pid/prgrp filter 
was fixed (bug #480252).

Debug info is more accurate in the face of symbols missing debug information 
(bug #484660).

A bug with C++ demangling has been fixed (bug #486753).
 
A workaround for a gcc 2.91.66 bug has been added (bug #481933).

A potentially serious bug with signals not being disabled was fixed (bug #491045).

A potentially serious bug fix with over-flowing the note buffer has been fixed. 
 
Some Pentium III systems were being identified incorrectly, and this has been fixed.

A small memory leak has been fixed.