2007-12-11 Philippe Elie <phil.el@wanadoo.fr>
* libutil++/op_bfd.h: remove sym_offset(), we can get what did
sym_offset() in a simpler way at call site
* libutil++/op_bfd.cpp:
* libpp/callgraph_container.cpp:
* libpp/profile_container.cpp:
* libpp/profile_container.h:
2007-11-25 Philippe Elie <phil.el@wanadoo.fr>
* daemon/opd_stats.c: dump invalid pc count, added to oprofile stats
fs 2.6.24-rc
2007-11-15 Dave Nomura <dcnltc@us.ibm.com>
* fixed a some typos in manually patched event files
2007-11-15 Dave Nomura <dcnltc@us.ibm.com>
* libpp/format_output.cpp: output XML SYMBOL_DATA for callers/callees
* libpp/format.h:
2007-11-15 Dave Nomura <dcnltc@us.ibm.com>
* events/ppc64/power6/event_mappings: updated Power6 event files
* events/ppc64/power6/events:
2007-11-15 Philippe Elie <phil.el@wanadoo.fr>
handle --root which act as a replacement (a prefix) for the / fs
* libpp/locate_images.cpp: root path search taking care if a
root path is given we never search through "/"
* libpp/locate_images.h:
* libpp/populate.cpp:
* libpp/populate_for_spu.cpp: use the right path to check last
modification time, this is yet again a change in behavior, module
was never found a this point, even if we found them later, and we
didn't try to check_mtime(), so opreport is again a bit more verbose
* libutil++/op_bfd.cpp: be more cautious when checking if we can try to
open the bfd file, don't open it in the '/' fs if a --root is given
* libpp/profile_spec.cpp: convey the root_path all over the code.
* libpp/profile_spec.h:
* pp/common_option.cpp:
* pp/common_option.h:
* pp/opannotate_options.cpp:
* pp/oparchive.cpp:
* pp/oparchive_options.cpp:
* pp/opgprof_options.cpp:
* pp/opreport_options.cpp:
* doc/opannotate.1.in: document --root
* doc/oparchive.1.in:
* doc/opgprof.1.in:
* doc/opreport.1.in:
* doc/oprofile.xml:
2007-11-10 Dave Nomura <dcnltc@us.ibm.com>
* libutil++/bfd_support.cpp: is_correct_function() is itended to deal
with missing debug information, if linenr != 0, don't call it since
we know we have valid debug info
2007-11-08 Haavard Skinnemoen <hskinnemoen@atmel.com>
* utils/opcontrol: Busybox's implementation of "kill" doesn't
understand the "-s SIG" option. Use "-SIG" instead.
2007-11-07 Philippe Elie <phil.el@wanadoo.fr>
Compile fix for gcc 2.95.3
* libpp/format_output.cpp: unused local var, then unused params
* libpp/format_output.h:
* libpp/op_header.cpp: missing include
* libutil++/bfd_spu_support.cpp: gcc 2.95.3 do not like struct bfd,
2.95 get it wrong probably but anyway all these struct are useless.
2007-11-06 Philippe Elie <phil.el@wanadoo.fr>
* pp/opannotate.cpp: for module we didn't use the right image name,
this bug appeared after 0.9.3 release
2007-11-05 Richard Purdie <rpurdie@openedhand.com>
* events/arm/armv6/events: Fix armv6 events to match mainline kernels
2007-11-04 Philippe Elie <phil.el@wanadoo.fr>
* utils/opcontrol: nmi watchdog is now cleanly handled in the kernel
but it can be on by default even if CONFIG_WATCHDOG=n, a sysctl
exists to put it off. We need to check this in opcontrol since we
returned success if a counter is missing and starting oprofile with
oprof_start silently failed.
2007-11-03 Richard Purdie <rpurdie@openedhand.com>
* pp/oparchive.cpp: Check list_files option before calling
create_dir() for debug files.
2007-11-03 Philippe Elie <phil.el@wanadoo.fr>
with --xml and --details, bfd open/close was done one time per
symbols, not one time per application. Solved by moving open/close
in the caller and cache bfd object, this work because --xml imply
symbols are sorted by app name then lib name. This is not perfect as
we can open/close multiple time bfd object with --separate=library
but still a real win. Quick test showed an improvement by over 40
times. Improvement grows as number of symbols per binary.
* libpp/xml_utils.cpp: move bfd open ...
* libpp/xml_utils.h:
* libpp/format_output.cpp: ... here
* libpp/format_output.h:
* pp/opreport.cpp:
2007-11-02 Philippe Elie <phil.el@wanadoo.fr>
* pp/opreport_options.cpp: --xml is not compatible with --sort since
xml output need some predefined sort options, we warned against
the use of --sort but didn't reset the sort options to default.
2007-11-01 Philippe Elie <phil.el@wanadoo.fr>
* pp/oparchive_options.cpp: oparchive need to be run as root if you
want to get save all binaries, generally I don't like to restrict
user to do silly things but --output-directory / is an easy typo
to do and will completely screw up the box.
2007-10-30 Philippe Elie <phil.el@wanadoo.fr>
* libpp/format_output.cpp: Fix #931882, xml output not changed at
the moment.
2007-10-30 Philippe Elie <phil.el@wanadoo.fr>
Prepare #931882 fix, continue to move down real filename
retrieval by extending image_name_storage which can hold now the
real image name and the image as derived from the sample filename.
Use the new api in a compatible way with the old behavior, no output
change intended.
* libpp/format_output.cpp:
* libpp/format_output.h:
* libpp/locate_images.cpp:
* libpp/locate_images.h:
* libpp/name_storage.cpp:
* libpp/name_storage.h:
* libpp/symbol.cpp:
* libpp/symbol.h:
* libpp/xml_utils.cpp:
* libpp/xml_utils.h:
* pp/opreport.cpp:
2007-10-29 Philippe Elie <phil.el@wanadoo.fr>
* libpp/image_errors.cpp: Pass to image_errors the archive_path through
an extra_images to ensure no duplicate error message occur nor we
miss error, this fix two bugs, one with differential profile
an error message was missing when two binary are missing, another
with opreport where the same error message was duplicate because in
on case we passed the real binary name (prefixed with archive_path)
while latter we passed the short form of the binary name.
* libpp/image_errors.h:
* libpp/callgraph_container.cpp:
* libpp/populate.cpp:
* libpp/populate_for_spu.cpp:
* libpp/xml_utils.cpp:
* libutil++/op_bfd.h:
* pp/opannotate.cpp: global var archive_path is no longer used
* pp/opannotate_options.cpp:
* pp/opannotate_options.h:
* pp/oparchive.cpp:
* pp/opgprof.cpp:
* pp/opgprof_options.h:
* pp/opreport.cpp:
* pp/opreport_options.cpp:
* pp/opreport_options.h:
2007-10-29 Philippe Elie <phil.el@wanadoo.fr>
Cleanup the way per spec archive_path is handled, as extra_images
is built using an archive_path it make no sense to pass archive_path
as parameter all over the place, rather make it a member of
extra_images and use it, anyway it was required the archive_path
passed to find_image_path() was the same as used to build the
extra_images.
As a side effect it fix a corner case: oparchive archive:tmp1 -o tmp2
was not working for module, archiving from an archive should work now.
* libpp/arrange_profiles.cpp:
* libpp/arrange_profiles.h:
* libpp/callgraph_container.cpp:
* libpp/callgraph_container.h:
* libpp/locate_images.cpp:
* libpp/locate_images.h:
* libpp/populate.cpp:
* libpp/populate.h:
* libpp/populate_for_spu.cpp:
* libpp/populate_for_spu.h:
* libpp/profile_spec.cpp:
* libpp/xml_utils.cpp:
* libpp/xml_utils.h:
* libutil++/op_bfd.cpp:
* libutil++/op_bfd.h:
* libutil++/op_spu_bfd.cpp:
* pp/opannotate.cpp:
* pp/oparchive.cpp:
* pp/oparchive_options.cpp:
* pp/oparchive_options.h:
* pp/opgprof.cpp:
* pp/opgprof_options.cpp:
* pp/opreport.cpp:
2007-10-29 Philippe Elie <phil.el@wanadoo.fr>
* libutil++/file_manip.cpp: do not create the output file if the input
file can not be read. It change oparchive behavior which created
empty file with input file like -rws--x--x 1 root root
2007-10-28 Philippe Elie <phil.el@wanadoo.fr>
* libpp/format_output.cpp: minor overkill
2007-10-27 Philippe Elie <phil.el@wanadoo.fr>
Fix #1819350, it turned out this implied another fix for differential
profile, extra_images must be built per profile specification not
globally so most of this patch pass an extra_images down to op_bfd.
The way image name resolution has changed. We fixup an image name
only when really needed, when opening a bfd file, when checking a
binary exists and when we try matching an image name with a profile
specification, this robustify differential profile. This patch also
allow a profile_spec ala image:*oprofile.ko which was broken too.
It's now mandatory to use the same -p option used with oparchive when
using an archive with other pp tools, it was already the case but was
broken, this patch does not clarify the documentation about this
point because I suspect oparchive should be modified to save at the
top level directory the contents of -p option, pp tools will use that
automatically and --image_path will be obsoleted when used with an
archive.
This patch except clarifying than --image-path with archive: is a bit
special does not add any user level visible change.
* libpp/arrange_profiles.cpp:
* libpp/arrange_profiles.h:
* libpp/callgraph_container.cpp:
* libpp/callgraph_container.h:
* libpp/locate_images.cpp:
* libpp/locate_images.h:
* libpp/populate.cpp:
* libpp/populate_for_spu.cpp:
* libpp/profile_container.cpp:
* libpp/profile_container.h:
* libpp/profile_spec.cpp:
* libpp/profile_spec.h:
* libpp/xml_utils.cpp:
* libpp/xml_utils.h:
* libutil++/Makefile.am:
* libutil++/op_bfd.cpp:
* libutil++/op_bfd.h:
* libutil++/op_spu_bfd.cpp:
* pp/common_option.cpp:
* pp/common_option.h:
* pp/opannotate.cpp:
* pp/opannotate_options.cpp:
* pp/oparchive.cpp:
* pp/oparchive_options.cpp:
* pp/opgprof.cpp:
* pp/opgprof_options.cpp:
* pp/opreport.cpp:
* pp/opreport_options.cpp:
2007-10-27 Philippe Elie <phil.el@wanadoo.fr>
* libpp/profile_spec.cpp: look like an obvious typo, we must fixup
the string we are parsing not a previously seen or empty string
2007-10-25 Philippe Elie <phil.el@wanadoo.fr>
* libutil++/sparse_array.h: size() return the max index while we
want the max index + 1, the only user is diff_container,
differential container is broken in 0.9.3. This fix bug #1820202
2007-10-25 Philippe Elie <phil.el@wanadoo.fr>
* utils/opcontrol: fix the previous opcontrol --reset patch, it was
broken if the daemon is running because --reset imply --dump so
we can use this --reset shortcut only if the daemon is not running.
2007-10-25 Philippe Elie <phil.el@wanadoo.fr>
* utils/opcontrol: if objdump is missing the error message is
obscure, this fix bug #1564920 by checking if objdump exists
and is an executable before using it
2007-10-25 Philippe Elie <phil.el@wanadoo.fr>
* utils/opcontrol: --reset doesn't need to load the module, it was
counter-intuitive than the sequence opcontrol --deinit;
opcontrol --reset must be followed by another opcontrol --deinit
2007-10-25 Philippe Elie <phil.el@wanadoo.fr>
* utils/opcontrol: error out earlier if the module is not loaded
even for --list-events or --dump to avoid obscure error message
2007-10-24 Philippe Elie <phil.el@wanadoo.fr>
* libop/op_config.h: bump sample filename format, should have been
done before 0.9.3
* libpp/op_header.cpp: do not use odb_open() to read the sample file
header, if it fails it can return EINVAL if sanity checking fails
so we can't know if the failure came from a sample file format
change or some other failure. This explain some "sample_filename:
Invalid argument" we got in bug report when user didn't cleanup
the sample file directory.
* libpp/profile.cpp: use read_header() first to check the sample file
format to get better error message.
2007-10-19 Philippe Elie <phil.el@wanadoo.fr>
* libop/op_events.c: force the use of hexa notation in event file for
the field intended to be in hexacidemal so make check will catch
attempt to use decimal notation for them.
* events/mips/20K/events: all event number were in decimal but parsed
as hexadecimal, this fix bug #1717298
* events/mips/24K/events:
* events/mips/25K/events:
* events/mips/5K/events:
* events/mips/r12000/events:
* events/mips/sb1/events:
* events/mips/vr5432/events:
* events/mips/vr5500/events:
* events/mips/34K/events: comment a bunch of overlapping event.
* events/ppc64/cell-be/unit_masks: well, 0 == 0x0 but it's easier
to catch real error by forcing hexa notation for default unit mask.
* libop/tests/alloc_counter_tests.c: fix mips/34K test, the used event
has been remoded
2007-10-17 Philippe Elie <phil.el@wanadoo.fr>
* doc/oprofile.xml: fix dead url
* utils/ophelp.c:
2007-10-17 Jason Yeh <jason.yeh@amd.com>
* events/x86-64/family10/events: This patch updates the events and
unit_masks files to match the BIOS and Kernel Developer's Guide
For AMD Family 10h Processors Rev 3.00 released on September 07,
2007. The notable changes are L3 cache events and updates to
corresponding unit masks.
* events/x86-64/family10/unit_masks:
2007-10-17 Philippe Elie <phil.el@wanadoo.fr>
* utils/opcontrol: newer shell accept if -z ... but older need
if test -z. This broke opcontrol --dump with bash 2.x and probably
other shell. With bash the error when running opcontrol as non root
was:
/usr/local/bin/opcontrol: line 1670: -z: command not found
2007-10-12 Philippe Elie <phil.el@wanadoo.fr>
* libop/op_events.c: check for '.' inside event name, they can't work
* events/i386/core_2/events: change '.' by '_' in events name
* events/ppc64/cell-be/events:
2007-10-11 Maynard Johnson <maynardj@us.ibm.com>
* events/ppc64/970MP/event_mappings: Turn off profiling in
hypervisor on 970MP to prevent lost interrupts
2007-10-10 Maynard Johnson <maynardj@us.ibm.com>
* events/ppc64/970MP/events:
* events/ppc64/970MP/event_mappings: Fix MMCR values
and counter-to-event mappings on a few 970MP groups
2007-10-09 Melchior Franz <mfranz@aon.at>
* libregex/op_regex.cpp: Fix broken string concatenation
2007-10-05 Will Cohen <wcohen@redhat.com>
* utils/opcontrol: Modify opcontrol to allow short form -l and -d for
normal users.
2007-09-25 Brian Twichell <tbrian@us.ibm.com>
* utils/opcontrol: Modify opcontrol to use a more inclusive
kernel range
2007-08-23 Maynard Johnson <maynardj@us.ibm.com>
* daemon/opd_anon.c: Make anonymous sample recording
work with 32-bit OProfile and 64-bit anonymous code
2007-08-14 John Levon <levon@movementarian.org>
* doc/oprofile.xml:
* doc/opcontrol.1.in: clarify behaviour of opcontrol -e
2007-08-02 Jens Wilke <jens.wilke@de.ibm.com>
* ChangeLog: Correct the date from previous patch (maynardj)
* libutil++/op_bfd.cpp: Fix sym_offset calculation.
Bug triggers on 64 bit systems with code mapped
with an address >4G.
2007-08-02 Jens Wilke <jens.wilke@de.ibm.com>
* libpp/profile_container.cpp: Remove unnecessary offset
calculation.
2007-07-18 Maynard Johnson <maynardj@us.ibm.com>
* doc/Makefile.am: Include buffers diagram in dist
* doc/oprofile.xml: Correct filename reference to opreport.xsd
2007-07-16 Maynard Johnson <maynardj@us.ibm.com>
* configure.in: Bump to 0.9.4cvs.
2007-07-16 Maynard Johnson <maynardj@us.ibm.com>
* configure.in:
* Makefile.am: Bump release to 0.9.3 and add a line to
EXTRA_DIST to distribute ChangeLog-2006.
2007-07-09 Maynard Johnson <maynardj@us.ibm.com>
* doc/opreport.xsd:
* libpp/xml_utils.cpp: Correct schema version bump
2007-07-02 Maynard Johnson <maynardj@us.ibm.com>
* libpp/populate_for_spu.cpp: Fix logic in is_spu_profile()
to handle '--separate=thread' option.
2007-06-27 Maynard Johnson <maynardj@us.ibm.com>
* ChangeLog: ChangeLog cleanup
2007-06-15 Maynard Johnson <maynardj@us.ibm.com>
* doc/opreport.xsd:
* libpp/xml_utils.cpp: Bump schema version to 2.0 to coincide with
callgraph elements added recently to the schema
2007-06-13 Maynard Johnson <maynardj@us.ibm.com>
* doc/oprofile.xml:
* events/Makefile.am:
* events/ppc64/power5++/event_mappings:
* events/ppc64/power5++/events:
* events/ppc64/power5++/unit_masks:
* libop/op_cpu_type.c:
* libop/op_cpu_type.h:
* libop/op_events.c:
* utils/opcontrol:
* utils/ophelp.c: Add support for POWER5++ (i.e., POWER5+ with PMU
set to POWER6 mode)
2007-06-12 dean gaudet <dean@arctic.org>
* events/i386/core_2/events:
* events/i386/core_2/unit_masks: updates to match
253669-022US ISA vol 3B from Intel. (Including fix
for #1580064 from Arun Sharma.)
2007-06-08 dean gaudet <dean@arctic.org>
* events/x86-64/hammer/events:
* events/x86-64/hammer/unit_masks: updates from December 2006
AMD update
2007-05-31 Maynard Johnson <maynardj@us.ibm.com>
* libpp/symbol.h:
* libutil++/bfd_support.h:
* m4/cellspubfdsupport.m4: Fixups for compile failures on older Linux
distributions
2007-05-29 Dave Nomura <dcnltc@us.ibm.com>
* libregex/demangle_symbol.cpp: Strip off leading '.' generated for elf6r-powerpc targets in mangled symbol names.
2007-05-24 Richard Purdie <rpurdie@openedhand.com>
* events/arm/xscale1/events:
* events/arm/xscale2/events: Add extra Xscale PMU event definitions
2007-05-24 Richard Purdie <rpurdie@openedhand.com>
* events/arm/armv6/events:
* events/arm/armv6/unit_masks:
* libop/op_cpu_type.c:
* libop/op_cpu_type.h:
* libop/op_events.c:
* utils/ophelp.c: Name ARM11 support to be consistent with the kernel,
remove some duplicate code and add some extra events
2007-05-24 Richard Purdie <rpurdie@openedhand.com>
* oprofile/pp/oparchive.cpp: If the debug files have the same name
as the original binary, the binary gets overwritten. Add in a .debug
directory to avoid this.
2007-05-23 Joerg Wagner <wagner@ccrl-nece.de>
* events/Makefile.am:
* events/arm/mpcore/events:
* events/arm/mpcore/unit_masks:
* libop/op_cpu_type.c:
* libop/op_cpu_type.h:
* libop/op_events.c:
* utils/ophelp.c: ARM MPCore support
2007-05-23 Riku Voipio <riku.voipio@iki.fi>
* events/Makefile.am:
* events/arm/armv6/events:
* events/arm/armv6/unit_masks:
* libop/op_cpu_type.c:
* libop/op_cpu_type.h:
* libop/op_events.c:
* utils/ophelp.c: ARM11 support
2007-05-21 Richard Purdie <rpurdie@openedhand.com>
* doc/opreport.xsd:
* libpp/format_output.cpp:
* libpp/format_output.h:
* libpp/xml_utils.cpp:
* libutil++/xml_output.cpp:
* libutil++/xml_output.h:
* pp/opreport.cpp:
* pp/opreport_options.cpp: Add callgraph XML output
2007-05-20 Dean Gaudet <dgaudet@users.sourceforge.net>
* events/i386/core_2/events: ITLB_MISS_RETIRED,MEM_LOAD_RETIRED
only work with counter 0
2007-05-20 Richard Purdie <rpurdie@openedhand.com>
* libpp/callgraph_container.cpp:
* libpp/callgraph_container.h:
* libpp/format_output.cpp:
* libpp/format_output.h:
* libpp/symbol.h:
* libpp/symbol_sort.cpp:
* libpp/symbol_sort.h:
* pp/opreport.cpp: Convert cg_collection to symbol_collection and use
dynamic casting allowing more code reuse for callgraphs
2007-05-19 Richard Purdie <rpurdie@openedhand.com>
* doc/oparchive.1.in:
* doc/oprofile.xml:
* pp/oparchive.cpp:
* pp/oparchive_options.cpp:
* pp/oparchive_options.h: Add --list-files option to list all files
that would be handled by oparchive.
2007-05-19 Olof Johansson <olof@lixom.net>
* events/Makefile.in:
* events/ppc64/pa6t/event_mappings:
* events/ppc64/pa6t/events:
* events/ppc64/pa6t/unit_masks:
* libop/op_cpu_type.c:
* libop/op_cpu_type.h:
* libop/op_events.c:
* utils/opcontrol:
* utils/ophelp.c: PA Semi 6T support
2007-05-09 Richard Purdie <rpurdie@openedhand.com>
* oprofile/daemon/opd_cookie.c: Fix syscall for ARM EABI
2007-05-10 Maynard Johnson <maynardj@us.ibm.com>
* libpp/Makefile.am:
* libpp/populate.cpp:
* libpp/populate_for_spu.cpp:
* libpp/populate_for_spu.h:
* libpp/profile.h:
* libpp/profile.cpp:
* libutil++/Makefile.am:
* libutil++/bfd_spu_support.cpp:
* libutil++/bfd_support.h:
* libutil++/op_bfd.h:
* libutil++/op_spu_bfd.cpp:
* libpp/xml_utils.cpp:
* libpp/profile_container.cpp:
* libpp/symbol.h: Patch 3 of 3 for adding support for profiling
Cell Broadband Engine SPU
2007-05-10 Maynard Johnson <maynardj@us.ibm.com>
* daemon/liblegacy/opd_sample_files.c:
* daemon/opd_events.c:
* daemon/opd_events.h:
* daemon/opd_interface.h:
* daemon/Makefile.am:
* daemon/opd_mangling.c:
* daemon/opd_sfile.c:
* daemon/opd_sfile.h:
* daemon/opd_spu.c:
* daemon/opd_trans.c:
* daemon/opd_trans.h:
* libop/op_sample_file.h: Patch 2 of 3 for adding support for profiling
Cell Broadband Engine SPU
2007-05-10 Maynard Johnson <maynardj@us.ibm.com>
* configure.in:
* doc/oprofile.xml:
* events/ppc64/cell-be/events:
* m4/Makefile.am:
* m4/cellspubfdsupport.m4:
* utils/opcontrol: Patch 1 of 3 for adding support for profiling
Cell Broadband Engine SPU
2007-04-25 Manoj Ekbote <manoj.ekbote@broadcom.com>
* events/mips/sb1/events: fix SB1 events
2007-04-19 Dave Nomura <dcnltc@us.ibm.com>
* events/ppc64/power6/events: counter 3 should have been used
in the specification of the default CYCLES event, not counter 1
2007-04-13 Will Cohen <wcohen@redhat.com>
* libop/op_cpu_type.c:
* libop/op_cpu_type.h: Move new entries to end of list
to improve backward compatibility with enum values.
2007-04-11 John Levon <levon@movementarian.org>
* doc/oprofile.xml: IRC channel is on OFTC
2007-04-10 Philippe Elie <phil.el@wanadoo.fr>
* utils/ophelp.c: if userspace doesn't recognize the cpu type
warn the user it can use timer mode or upgrade oprofile.
2007-04-04 Maynard johnson <maynardj@us.ibm.com>
* libutil++/file_manip.cpp: catch result from chown() to avoid
compilation warning (treated as error)
2007-03-23 Jason Yeh <jason.yeh@amd.com>
* events/Makefile.am:
* events/x86-64/family10/events:
* events/x86-64/family10/unit_masks:
* libop/op_cpu_type.c:
* libop/op_cpu_type.h:
* libop/op_events.c:
* utils/ophelp.c: Add AMD Family 10 support.
2007-02-28 Rob Bradford <rob@o-hand.com>
* pp/oparchive.cpp: return from main
2007-02-21 Rob Bradford <rob@o-hand.com>
Reverted previous patches.
* pp/oparchive.cpp: Save debuginfo files in .debug sub directory.
* pp/oparchive.cpp:
* pp/oparchive_options.h:
* pp/oparchive_options.cpp: List files option.
2007-02-21 Rob Bradford <rob@o-hand.com>
* pp/oparchive.cpp: Save debuginfo files in .debug sub directory.
* pp/oparchive.cpp:
* pp/oparchive_options.h:
* pp/oparchive_options.cpp: List files option.
2007-02-16 Philippe Elie <phil.el@wanadoo.fr>
* daemon/opd_mangling.c: missing initialisation when mangling an
anon and callgraph filename. Fix given by Amitabha Roy.
2007-02-06 Dave Nomura <dcnltc@us.ibm.com>
* libpp/symbol.h:
* libutil++/Makefile.am:
* libutil++/sparse_array.h: represent count_array_t as a sparse
array type based on std::map rather than growable_vector which is
based on std::vector to reduce the excessive heap usage on large
profiles.
2007-02-02 Amitabha Roy <amitabha.roy@gmail.com>
* daemon/opd_anon.c:
* daemon/opd_anon.h:
* daemon/opd_mangling.c:
* libop/op_mangle.c:
* libop/op_mangle.h:
* libpp/parse_filename.cpp: save and report name of anonymous
mapping if there is one (as self-contradictory as that
sounds).
2007-02-02 Dave Nomura <dcnltc@us.ibm.com>
* events/Makefile.am:
* events/ppc64/970MP/event_mappings:
* events/ppc64/970MP/events:
* events/ppc64/970MP/unit_masks:
* libop/op_cpu_type.c:
* libop/op_cpu_type.h:
* libop/op_events.c:
* utils/ophelp.c: PPC970MP has different hardware counters than the rest
of the PPC family and must be treated as a different architecture
by oprofile.
2007-01-31 Dave Nomura <dcnltc@us.ibm.com>
* libpp/xml_utils.cpp: opreport -X was still getting some incorrect
symbols attributed to the <binary> when processing a --separate=lib
profile.
2007-01-29 Philippe Elie <phil.el@wanadoo.fr>
* daemon/oprofiled.c: no need for 0755 for the log file mode, use 0644
* utils/opcontrol: don't create the log in opcontrol
2007-01-26 Dave Nomura <dcnltc@us.ibm.com>
* libpp/format_output.cpp:
* libpp/xml_utils.cpp: opreport -X was mixing the symbols associated with
an application with those of a library when processing a --separate=lib
profile.
2007-01-02 Dave Nomura <dcnltc@us.ibm.com>
* events/Makefile.am:
* events/ppc64/power6/event_mappings:
* events/ppc64/power6/events:
* events/ppc64/power6/unit_masks:
* libop/op_cpu_type.c:
* libop/op_cpu_type.h:
* libop/op_events.c:
* utils/opcontrol:
* utils/ophelp.c: Add support for Power6
See ChangeLog-2006 for earlier changelogs.