Static build failure with binutils > 2.40
oprofile: fix build on gcc-12
install docs wrong debian has libiberty
I overlooked the previous configure setting.
Hello William, Here is an extended snippet of code from configure.ac (including my changes): dateopt='' if test -n "$SOURCE_DATE_EPOCH" ; then if date --version | grep -q GNU ; then dateopt="-u -d @$SOURCE_DATE_EPOCH" else dateopt="-u -r $SOURCE_DATE_EPOCH" fi fi dnl for the man page DATE="`date $dateopt '+%a %d %B %Y'`" AC_DEFINE_UNQUOTED([BUILD_DATE], ["$DATE"], [Use reproducible build date]) AC_SUBST(DATE) As you can see, the purpose of the dateopt variable is precisely to pass the correct parameters...
Remove non-Posix use of FTW_ACTIONRETVAL
Going to close this because oprofile, commit 5ed9bda8baf7465172a99ff86ed7f46397b06c7f addresses this.
The limitation to this approach is that builds are only repeatable if they are built in the same day. Maybe something like https://reproducible-builds.org/docs/source-date-epoch/ would provide more reproducible builds. Also reproducible is spelled as "reproducidle" in 0e4969.
I forgot to remove the path on the last two code excerps. Would it be possible for an admin to edit the post to remove it for me please?
Error when compiled with Cray compiler
Sorry. I didn't see this pull request until now. A similar patch was submitted and has been merged into oprofile, commit 5ed9bda8baf7465172a99ff86ed7f46397b06c7f .
Fix build with musl
Fix configure when /bin/sh is not bash
Bump version to 1.5.0git
Do you have some feedback on this merge request?
Update relase checklist.
Add release notes for oprofile-1.4.0
Add news and download inform for oprofile-1.4.0.
Update the documentation information to include thunderx2.
ocount leaves orphan process on error
Does not build with binutils-gdb master.
Change version to 1.4.0 in preparation for GA
Include op_bfd_wrappers.h in the distribution
Does not build with binutils-gdb master.
Hi, Tried this patch and the package compiled well. I will send this patch to Buildroot. Thanks you for fix!
master as of now compiles fine for me with binutils 2.34 and 2.33.
Could you try the current master branch of oprofile and verify that this fixes the reported problem?
There are still a number of distributions that have older bersions of binutils. This needs to be conditional so oprofile can be compiled with the older binutils. Should have a set of wrapper defines in another header file and have a a test in configure.ac to select the correct wrappers. I have put a patch that does that on https://sourceforge.net/p/oprofile/oprofile/ci/bfd234/tree/ The patch allow oprofile to build on fedora 30 (binutils 2.31) and fedora rawhide (binutils 2.34). If the patch looks...
Macro wrappers to handle the binutils 2.34 api changes
This patch fixes the build for me using binutils 2.34. It will cause compile failures for previous versions, so it needs to be made conditional if compatibility with older versions is needed.
oprofile-tests Add test to exercise libjvmti_oprofile.so
Does not build with binutils-gdb master.
Use the mmap offset to correctly compute the IP location in a file
What does the strace output look like for the oprofile-1.1.0 operf? Does the oprofile-1.1.0 build have any patches in it? Or is it the completely stock oprofile-1.1.0?
Hi William, Yes I too grep from the same and found none, but I could see the profiling works good when the version was downgraded to 1.1.0. Just added the profiling tools and oprofile from meta-Openembedded in yocto & selected oprofile from Buildroot. Found the same issue. Here is the log; root@beaglebone-yocto:~# strace operf --version execve("/usr/bin/operf", ["operf", "--version"], 0xbec3ad94 /* 15 vars */) = 0 brk(NULL) = 0x50c000 uname({sysname="Linux", nodename="beaglebone-yocto", ...}) = 0...
Hi, I too grep for the same and didn't get any way to trace this. I just build the package from Yocto (meta-openembedded) and buildroot both, which yields the same. If you go back to installing oprofile-1.1.0 everything works fine? Yes with the same build system for 1.1.0 was working good. Find the attached strace log and which operf
It doesn't look like "Operation not support" is coming from operf. The message doesn't seem to be in the oprofile source code. could you do a "which operf" to check to see which operf is being run? Could you give more destails on how oprofile-1.3.0 was built and installed on the machine? Where is oprofile installed on the machine? You might try to use the full path to make sure that the desired one is being used. Is it possible run "strace operf --version" to get a better idea when and where things...
Oprofile fails to run on ARM iMX.6 boards
Proposed fix for oprofile JIT bad event spec error
Use https rather http in links on start page.
On Thu, 2019-07-25 at 14:39 +0000, William Cohen wrote: status: open --> open-fixed Group: --> Comment: I massaged the patch as it wasn't formatted quite right and didn't apply with a "git am". The logic seems reasonable and I verified that the patch fixed the stated problem. It is now in the upstrem oprofile git repository. Thanks. Will: Thanks. It was generated via git gui so not sure why it wouldn't apply with git am. Carl Love
Patch was committed by Will Cohen. The commit entry is: commit 8aacb573cd2c4a2f8160d99ff100ad0aa5e7859d Author: Carl Love cel@us.ibm.com Date: Thu Jul 25 10:24:16 2019 -0400 Only start the application if the perf events setup was successful Changes the order of starting the application and performance events. Given this change we have a new issue. The issue is the routine start_counting() calls fork, creating app_PID process. The parent then tries to setup the performance events, then if the performance...
ocount leaves orphan process on error
I massaged the patch as it wasn't formatted quite right and didn't apply with a "git am". The logic seems reasonable and I verified that the patch fixed the stated problem. It is now in the upstrem oprofile git repository. Thanks.
Only start the application if the perf events setup was successful
The second patch can be used to force the error seen by the user. It is for testing purposes only and is not be committted to the repository. Will, let me know if you can read the patches OK or not. Thanks.
Per Will's request, I am attaching the two patches separately. The first patch is the fix for the issue.
Could you separate the the test and the actual fix in the patch into two separate patches? Right now they are all combined and done want that for the actual commit. The logic looks reasonable but would like to test thing a bit with the split patches.
ophelp: missing "triggers" in some Power8 events
I have attached the message/patch sent to the mailing list.
ocount leaves orphan process on error
I will happily close this once I figure out how to do so. (??)
I'm not able to reproduce, either. Closing...
The first issue noted was perf failed to start with "ocount --system-wide -e PM_L1MISS_LAT_EXC_32". Tested on a Power 8 system: /pe_counting# ocount --system-wide -e PM_L1MISS_LAT_EXC_32 ocount: Press Ctl-c or 'kill -SIGINT 122560' to stop counting ^C Events were actively counted for 3.3 seconds. Event counts (actual) for the whole system: Event Count % time counted PM_L1MISS_LAT_EXC_32 101,107 100.00 That does give non zero results, no perf errors. I tried the 256, 1024 and 2048 events as well and...
Power8: a few event misspellings with ophelp
Fixed by the following patch that is in oprofile-1.2.0: commit 6fcd5aa57482a58fcb0166982fed517fbf7040fb Author: Carl E. Love cel@us.ibm.com Date: Thu Mar 17 13:49:41 2016 -0700 POWER 8 processor event spelling fixes Will: Here is a patch to fix the spelling errors in the Power 8 events. See OProfile bugzilla number 281. This patch corrects the spelling errors. Carl Love ------------------------------------------------------------ POWER 8 processor event spelling fixes. Fixed the spelling of six of...
remove unsupported POWER8 PMU events from OProfile/libpfm4
I believe this issue can be closed.
I believe this issue can be closed?
Remove non-Posix use of FTW_ACTIONRETVAL
Oprofile-tests Hygon Dhyana CPU support
Add support for Hygon Dhyana CPU
oprofile - fix compiler warnings seen with newer gcc
oprofile: add support for Cavium ThunderX2
Handle empty event name spec gracefully for PPC
Bump version to 1.4.0git
Add Intel Goldmont Plus link.
No manpage for opjitconv binary
Correct version number to 1.3.0.
Add release notes for oprofile-1.3.0
Check in modified and new event list files for OProfile-1.3.0
Change version to 1.3.0 in preparation for GA
A man for opjitconv added with git commit 51f2b9495971
Eliminate possible string truncations and buffer overflows in opjitconv.c
Rework string handling in op_pe_utils.cpp to avoid truncation and overflows
Make it clear to GCC that the strncpy limited to the buffer size
Remove problem typecasting in option_base::option_base
Make the always_true function signature match its use
oprofile-tests increase iterations for anon_remap test
doc/ophelp.xsd: fix ophelp validation schema (event's attribute "ext")
Initialize the trans structure fields to values unused by perf samples
PowerPC: Remove trailing comma in cpu_name.
oprofile: Add support for Intel Goldmont Plus (Geminilake)
yes, the gcc patch is working nicely for us there, so no further effort required.
According to https://wiki.debian.org/ReproducibleBuilds/TimestampsFromCPPMacros The DATE and TIME defines are already addressed by a patch to gcc.
A similar change is needed in libutil/op_version.c replacing the c preprocessor DATE and TIME defines. There was some earlier discussion on the mailing list on this, should revise that work to use the DATE from the configure.
make build reproducible
opjitconv: add initial manpage
Fix macro in operf man page
power9, remove event entries with 0x00 values.
Fix typos in op_netburst.c file
No manpage for opjitconv binary
Bump version to 1.3.0git