From: Maynard J. <may...@us...> - 2008-05-21 21:53:04
|
We are pleased to announce OProfile 0.9.4 Release Candidate 1. You can download this release at: http://sourceforge.net/project/showfiles.php?group_id=16191 This release has many new features and fixes. The biggest change with this release is the addition of support for profiling JITed code. Java agents that support Java 1.4 (JVMPI) and later (JVMTI) are included with OProfile. See the release notes below for more details. Please test this release candidate and let us know (good or bad) how it works. Thanks. -Maynard Johnson ------------------------------------------------------------------------------ Release Notes =============== OProfile is a powerful system-wide profiler for Linux. Read more at http://oprofile.sf.net OProfile 0.9.4 has been released. OProfile is still in alpha, but has been proven stable for many users. New features ------------ OProfile now supports profiling Java applications. See Section 1.4 "Installation" in the user manual for instructions on how to build OProfile so that it includes this support. OProfile also includes a framework for adding support for profiling other just-in-time (JIT) compiled languages. A new manual titled "OProfile JIT agent developer guide" is provided to aid developers. Added Xen support for IA64 Added AVR32 support Add '--root' option to opreport which act as a replacement (a prefix) for the / fs Performance improvement for 'opreport --xml --details'. OProfile tends to output obscure error message. Some of these have been fixed. Changes required to build with gcc 4.3 Bug fixes --------- Fix sym_offset calculation bug, seen on 64 bit systems with code mapped with an address >4G Fix opcontrol's calculation of kernel address range to include code found in sections other than just .text Fix opcontrol's short forms of --list-events (-l) and --dump (-d) so they work for non-root users Fix MMCR values and counter-to-event mappings on a few 970MP groups Turn off profiling in hypervisor on 970MP to prevent lost interrupts Update family10 events and unit_masks files to match the BIOS and Kernel Developer's Guide (current, as of May 20, 2008) Since --xml is not compatible with the --sort option, we warned against the use of --sort, but didn't reset the sort options to default. This is now fixed. Fix silent failure of oprof_start for when a counter is missing (e.g, NMI watchdog is up) Change opcontrol to use "-SIG" instead of "-s SIG" since Busybox's implementation of "kill" doesn't understand the "-s SIG" option Fix 'opannotate -s' to work with inlined code Update POWER6 event files (add new event groupings; make some fixes to others) Fix "Dangling ESCAPE CODE" error that can occur on Cell BE SPE profiling Fix opreport for Cell BE SPE profiles to attribute samples to dynamically generated call stubs executed from SPE stack (for example, I/O calls to the libc library) Fix loop in 'opcontrol --dump' code when using --session-dir on a network drive (clock issues) Fix problem with 'oparchive --list-files' that can occur if file doesn't exist Fix user/kernel domain profiling switches for ppc64 architectures Fix the bfd_get_synthetic_symtab check in m4 macro to work correctly with '--with-binutils' configure option Fix ARM big-endian syscall bug #1820202, differential profile broken in 0.9.3, is fixed bug #1717298, mips events have incorrect id numbers, is fixed, all event number > 9 was incorrectly setup bug #1564920, opcontrol does not check if objdump exists is fixed, we error out and pinpoint the right error now bug #1819350, oparchive doesn't work with kernel module, is fixed (need another related fix, if you read that I forget about that, ping me please -- phe) bug #1828566, --xml and -t and callgraph now output symboldata for all caller/callees bug #1930788, opreport error: basic_string::erase Known problems -------------- When using callgraph profiling, it's possible that invalid sample files are created (bug #1685267). Many Alpha ev67 events do not work (bug #931875). A few Pentium IV events are not supported (bug #841099). For 2.2 kernels, the module must be compiled as the same user that owns the kernel source tree. With an AMD64 kernel, OProfile must be built in 64 bit mode due to lack of kernel support. opreport -c gives strange output for binaries without symbols. |
From: Dave N. <dc...@us...> - 2008-05-28 16:09:56
|
I installed RC4 on a SLES9 PPC970MP machine and Oprofile passed our internal oprofile test suite. Maynard Johnson wrote: > We are pleased to announce OProfile 0.9.4 Release Candidate 1. You can download > this release at: > http://sourceforge.net/project/showfiles.php?group_id=16191 > > This release has many new features and fixes. The biggest change with this > release is the addition of support for profiling JITed code. Java agents that > support Java 1.4 (JVMPI) and later (JVMTI) are included with OProfile. See the > release notes below for more details. > > Please test this release candidate and let us know (good or bad) how it works. > > Thanks. > -Maynard Johnson > ------------------------------------------------------------------------------ > > > Release Notes > =============== > OProfile is a powerful system-wide profiler for Linux. Read > more at http://oprofile.sf.net > > OProfile 0.9.4 has been released. OProfile is still in alpha, > but has been proven stable for many users. > > New features > ------------ > > OProfile now supports profiling Java applications. See Section 1.4 > "Installation" in the user manual for instructions on how to build > OProfile so that it includes this support. > > OProfile also includes a framework for adding support for profiling > other just-in-time (JIT) compiled languages. A new manual titled > "OProfile JIT agent developer guide" is provided to aid developers. > > Added Xen support for IA64 > > Added AVR32 support > > Add '--root' option to opreport which act as a replacement (a prefix) > for the / fs > > Performance improvement for 'opreport --xml --details'. > > OProfile tends to output obscure error message. Some of these have been fixed. > > Changes required to build with gcc 4.3 > > Bug fixes > --------- > > Fix sym_offset calculation bug, seen on 64 bit systems with code mapped > with an address >4G > > Fix opcontrol's calculation of kernel address range to include code found > in sections other than just .text > > Fix opcontrol's short forms of --list-events (-l) and --dump (-d) so they > work for non-root users > > Fix MMCR values and counter-to-event mappings on a few 970MP groups > > Turn off profiling in hypervisor on 970MP to prevent lost interrupts > > Update family10 events and unit_masks files to match the BIOS and Kernel > Developer's Guide (current, as of May 20, 2008) > > Since --xml is not compatible with the --sort option, we warned against > the use of --sort, but didn't reset the sort options to default. This is > now fixed. > > Fix silent failure of oprof_start for when a counter is missing (e.g, NMI > watchdog is up) > > Change opcontrol to use "-SIG" instead of "-s SIG" since Busybox's > implementation of "kill" doesn't understand the "-s SIG" option > > Fix 'opannotate -s' to work with inlined code > > Update POWER6 event files (add new event groupings; make some fixes to others) > > Fix "Dangling ESCAPE CODE" error that can occur on Cell BE SPE profiling > > Fix opreport for Cell BE SPE profiles to attribute samples to dynamically > generated call stubs executed from SPE stack (for example, I/O calls to > the libc library) > > Fix loop in 'opcontrol --dump' code when using --session-dir on a network > drive (clock issues) > > Fix problem with 'oparchive --list-files' that can occur if file doesn't exist > > Fix user/kernel domain profiling switches for ppc64 architectures > > Fix the bfd_get_synthetic_symtab check in m4 macro to work correctly with > '--with-binutils' configure option > > Fix ARM big-endian syscall > > bug #1820202, differential profile broken in 0.9.3, is fixed > > bug #1717298, mips events have incorrect id numbers, is fixed, all event > number > 9 was incorrectly setup > > bug #1564920, opcontrol does not check if objdump exists is fixed, we error > out and pinpoint the right error now > > bug #1819350, oparchive doesn't work with kernel module, is fixed (need another > related fix, if you read that I forget about that, ping me please -- phe) > > bug #1828566, --xml and -t and callgraph now output symboldata for all > caller/callees > > bug #1930788, opreport error: basic_string::erase > > > Known problems > -------------- > > When using callgraph profiling, it's possible that invalid sample > files are created (bug #1685267). > > Many Alpha ev67 events do not work (bug #931875). > > A few Pentium IV events are not supported (bug #841099). > > For 2.2 kernels, the module must be compiled as the same user > that owns the kernel source tree. > > With an AMD64 kernel, OProfile must be built in 64 bit mode due to lack > of kernel support. > > opreport -c gives strange output for binaries without symbols. > > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2008. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > _______________________________________________ > oprofile-list mailing list > opr...@li... > https://lists.sourceforge.net/lists/listinfo/oprofile-list > |
From: Daniel H. <dan...@li...> - 2008-05-29 15:40:50
|
Hi Maynard and all others, I've tested the RC1 on a IBM System z 64bit platform (aka s390x) with SLES 10.1 installed. My tests were successfully finished. So I give an OK for this release candidate. Kind regards, Daniel Hansel Maynard Johnson wrote: > We are pleased to announce OProfile 0.9.4 Release Candidate 1. You can download > this release at: > http://sourceforge.net/project/showfiles.php?group_id=16191 > > This release has many new features and fixes. The biggest change with this > release is the addition of support for profiling JITed code. Java agents that > support Java 1.4 (JVMPI) and later (JVMTI) are included with OProfile. See the > release notes below for more details. > > Please test this release candidate and let us know (good or bad) how it works. > > Thanks. > -Maynard Johnson > ------------------------------------------------------------------------------ > > > Release Notes > =============== > OProfile is a powerful system-wide profiler for Linux. Read > more at http://oprofile.sf.net > > OProfile 0.9.4 has been released. OProfile is still in alpha, > but has been proven stable for many users. > > New features > ------------ > > OProfile now supports profiling Java applications. See Section 1.4 > "Installation" in the user manual for instructions on how to build > OProfile so that it includes this support. > > OProfile also includes a framework for adding support for profiling > other just-in-time (JIT) compiled languages. A new manual titled > "OProfile JIT agent developer guide" is provided to aid developers. > > Added Xen support for IA64 > > Added AVR32 support > > Add '--root' option to opreport which act as a replacement (a prefix) > for the / fs > > Performance improvement for 'opreport --xml --details'. > > OProfile tends to output obscure error message. Some of these have been fixed. > > Changes required to build with gcc 4.3 > > Bug fixes > --------- > > Fix sym_offset calculation bug, seen on 64 bit systems with code mapped > with an address >4G > > Fix opcontrol's calculation of kernel address range to include code found > in sections other than just .text > > Fix opcontrol's short forms of --list-events (-l) and --dump (-d) so they > work for non-root users > > Fix MMCR values and counter-to-event mappings on a few 970MP groups > > Turn off profiling in hypervisor on 970MP to prevent lost interrupts > > Update family10 events and unit_masks files to match the BIOS and Kernel > Developer's Guide (current, as of May 20, 2008) > > Since --xml is not compatible with the --sort option, we warned against > the use of --sort, but didn't reset the sort options to default. This is > now fixed. > > Fix silent failure of oprof_start for when a counter is missing (e.g, NMI > watchdog is up) > > Change opcontrol to use "-SIG" instead of "-s SIG" since Busybox's > implementation of "kill" doesn't understand the "-s SIG" option > > Fix 'opannotate -s' to work with inlined code > > Update POWER6 event files (add new event groupings; make some fixes to others) > > Fix "Dangling ESCAPE CODE" error that can occur on Cell BE SPE profiling > > Fix opreport for Cell BE SPE profiles to attribute samples to dynamically > generated call stubs executed from SPE stack (for example, I/O calls to > the libc library) > > Fix loop in 'opcontrol --dump' code when using --session-dir on a network > drive (clock issues) > > Fix problem with 'oparchive --list-files' that can occur if file doesn't exist > > Fix user/kernel domain profiling switches for ppc64 architectures > > Fix the bfd_get_synthetic_symtab check in m4 macro to work correctly with > '--with-binutils' configure option > > Fix ARM big-endian syscall > > bug #1820202, differential profile broken in 0.9.3, is fixed > > bug #1717298, mips events have incorrect id numbers, is fixed, all event > number > 9 was incorrectly setup > > bug #1564920, opcontrol does not check if objdump exists is fixed, we error > out and pinpoint the right error now > > bug #1819350, oparchive doesn't work with kernel module, is fixed (need another > related fix, if you read that I forget about that, ping me please -- phe) > > bug #1828566, --xml and -t and callgraph now output symboldata for all > caller/callees > > bug #1930788, opreport error: basic_string::erase > > > Known problems > -------------- > > When using callgraph profiling, it's possible that invalid sample > files are created (bug #1685267). > > Many Alpha ev67 events do not work (bug #931875). > > A few Pentium IV events are not supported (bug #841099). > > For 2.2 kernels, the module must be compiled as the same user > that owns the kernel source tree. > > With an AMD64 kernel, OProfile must be built in 64 bit mode due to lack > of kernel support. > > opreport -c gives strange output for binaries without symbols. > > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2008. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > _______________________________________________ > oprofile-list mailing list > opr...@li... > https://lists.sourceforge.net/lists/listinfo/oprofile-list |
From: William C. <wc...@nc...> - 2008-06-01 17:52:29
|
Maynard Johnson wrote: > We are pleased to announce OProfile 0.9.4 Release Candidate 1. You can download > this release at: > http://sourceforge.net/project/showfiles.php?group_id=16191 > > This release has many new features and fixes. The biggest change with this > release is the addition of support for profiling JITed code. Java agents that > support Java 1.4 (JVMPI) and later (JVMTI) are included with OProfile. See the > release notes below for more details. > > Please test this release candidate and let us know (good or bad) how it works. > > Thanks. > -Maynard Johnson Hi Maynard, Thanks so much for working on a releasing Oprofile 0.9.4. What is the logic for having oprofile require a special user and group oprofile for installation? On rpm builds the build machine might not have the special user and groups oprofile this may present an issue. Noticed that the Fedora 9 kernel (2.6.25.3-18.fc9) on x86_64 started with "nmi_watchdog=2" doesn't work with oprofile because performance counter 0 is used for the watchdog. There were some patches submitted earlier earlier that allowed the performance counter allocation to occur even when some performance counter are already reserved. Would it be possible to roll this patch into 0.9.4? Below is a link to the related email. http://marc.info/?l=oprofile-list&m=120716521708920&w=2 Will > ------------------------------------------------------------------------------ > > > Release Notes > =============== > OProfile is a powerful system-wide profiler for Linux. Read > more at http://oprofile.sf.net > > OProfile 0.9.4 has been released. OProfile is still in alpha, > but has been proven stable for many users. > > New features > ------------ > > OProfile now supports profiling Java applications. See Section 1.4 > "Installation" in the user manual for instructions on how to build > OProfile so that it includes this support. > > OProfile also includes a framework for adding support for profiling > other just-in-time (JIT) compiled languages. A new manual titled > "OProfile JIT agent developer guide" is provided to aid developers. > > Added Xen support for IA64 > > Added AVR32 support > > Add '--root' option to opreport which act as a replacement (a prefix) > for the / fs > > Performance improvement for 'opreport --xml --details'. > > OProfile tends to output obscure error message. Some of these have been fixed. > > Changes required to build with gcc 4.3 > > Bug fixes > --------- > > Fix sym_offset calculation bug, seen on 64 bit systems with code mapped > with an address >4G > > Fix opcontrol's calculation of kernel address range to include code found > in sections other than just .text > > Fix opcontrol's short forms of --list-events (-l) and --dump (-d) so they > work for non-root users > > Fix MMCR values and counter-to-event mappings on a few 970MP groups > > Turn off profiling in hypervisor on 970MP to prevent lost interrupts > > Update family10 events and unit_masks files to match the BIOS and Kernel > Developer's Guide (current, as of May 20, 2008) > > Since --xml is not compatible with the --sort option, we warned against > the use of --sort, but didn't reset the sort options to default. This is > now fixed. > > Fix silent failure of oprof_start for when a counter is missing (e.g, NMI > watchdog is up) > > Change opcontrol to use "-SIG" instead of "-s SIG" since Busybox's > implementation of "kill" doesn't understand the "-s SIG" option > > Fix 'opannotate -s' to work with inlined code > > Update POWER6 event files (add new event groupings; make some fixes to others) > > Fix "Dangling ESCAPE CODE" error that can occur on Cell BE SPE profiling > > Fix opreport for Cell BE SPE profiles to attribute samples to dynamically > generated call stubs executed from SPE stack (for example, I/O calls to > the libc library) > > Fix loop in 'opcontrol --dump' code when using --session-dir on a network > drive (clock issues) > > Fix problem with 'oparchive --list-files' that can occur if file doesn't exist > > Fix user/kernel domain profiling switches for ppc64 architectures > > Fix the bfd_get_synthetic_symtab check in m4 macro to work correctly with > '--with-binutils' configure option > > Fix ARM big-endian syscall > > bug #1820202, differential profile broken in 0.9.3, is fixed > > bug #1717298, mips events have incorrect id numbers, is fixed, all event > number > 9 was incorrectly setup > > bug #1564920, opcontrol does not check if objdump exists is fixed, we error > out and pinpoint the right error now > > bug #1819350, oparchive doesn't work with kernel module, is fixed (need another > related fix, if you read that I forget about that, ping me please -- phe) > > bug #1828566, --xml and -t and callgraph now output symboldata for all > caller/callees > > bug #1930788, opreport error: basic_string::erase > > > Known problems > -------------- > > When using callgraph profiling, it's possible that invalid sample > files are created (bug #1685267). > > Many Alpha ev67 events do not work (bug #931875). > > A few Pentium IV events are not supported (bug #841099). > > For 2.2 kernels, the module must be compiled as the same user > that owns the kernel source tree. > > With an AMD64 kernel, OProfile must be built in 64 bit mode due to lack > of kernel support. > > opreport -c gives strange output for binaries without symbols. > > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2008. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > _______________________________________________ > oprofile-list mailing list > opr...@li... > https://lists.sourceforge.net/lists/listinfo/oprofile-list > |
From: Daniel H. <dan...@li...> - 2008-06-02 09:16:00
|
William Cohen wrote: > Hi Maynard, > > Thanks so much for working on a releasing Oprofile 0.9.4. > > What is the logic for having oprofile require a special user and group oprofile > for installation? On rpm builds the build machine might not have the special > user and groups oprofile this may present an issue. > Hi William, the special user account is introduced due to fact that the handling with potentially insecure data (i.e. the conversion of dump file produced during JIT profiling) may "offer" a point to be attacked (e.g. buffer overflow, etc.). Therefore we decided to run the conversion step using a special user account. This mechanism is commonly used for such things. > Noticed that the Fedora 9 kernel (2.6.25.3-18.fc9) on x86_64 started with > "nmi_watchdog=2" doesn't work with oprofile because performance counter 0 is > used for the watchdog. There were some patches submitted earlier earlier that > allowed the performance counter allocation to occur even when some performance > counter are already reserved. Would it be possible to roll this patch into > 0.9.4? Below is a link to the related email. > > http://marc.info/?l=oprofile-list&m=120716521708920&w=2 Maybe Maynard could answer this question. > > Will > [snip] Kind regards, Daniel |
From: Maynard J. <may...@us...> - 2008-06-02 12:37:30
|
William Cohen wrote: > Maynard Johnson wrote: > >> We are pleased to announce OProfile 0.9.4 Release Candidate 1. You can download >> this release at: >> http://sourceforge.net/project/showfiles.php?group_id=16191 >> >> This release has many new features and fixes. The biggest change with this >> release is the addition of support for profiling JITed code. Java agents that >> support Java 1.4 (JVMPI) and later (JVMTI) are included with OProfile. See the >> release notes below for more details. >> >> Please test this release candidate and let us know (good or bad) how it works. >> >> Thanks. >> -Maynard Johnson >> > > Hi Maynard, > > Thanks so much for working on a releasing Oprofile 0.9.4. > > What is the logic for having oprofile require a special user and group oprofile > for installation? On rpm builds the build machine might not have the special > user and groups oprofile this may present an issue. > One thing to add to Daniel's reply . . . The configure script puts out a warning message if the 'oprofile' user and group accounts do not exist, and 'make install' will fail if they do not exist. Will this be a problem for your rpm build machines? Please look at the new file in oprofile's root source directory called README_PACKAGERS for additional info on spec file changes that are needed to support these required user and group accounts and let me know if you see any problems with this. > Noticed that the Fedora 9 kernel (2.6.25.3-18.fc9) on x86_64 started with > "nmi_watchdog=2" doesn't work with oprofile because performance counter 0 is > used for the watchdog. There were some patches submitted earlier earlier that > allowed the performance counter allocation to occur even when some performance > counter are already reserved. Would it be possible to roll this patch into > 0.9.4? Below is a link to the related email. > > http://marc.info/?l=oprofile-list&m=120716521708920&w=2 > Will, it looks to me like John gave his approval for that patch on April 16. Assuming that's the case, once you commit it, I can spin a release candidate 2 if you can test it. -Maynard > Will > > >> ------------------------------------------------------------------------------ >> >> >> Release Notes >> =============== >> OProfile is a powerful system-wide profiler for Linux. Read >> more at http://oprofile.sf.net >> >> OProfile 0.9.4 has been released. OProfile is still in alpha, >> but has been proven stable for many users. >> >> New features >> ------------ >> >> OProfile now supports profiling Java applications. See Section 1.4 >> "Installation" in the user manual for instructions on how to build >> OProfile so that it includes this support. >> >> OProfile also includes a framework for adding support for profiling >> other just-in-time (JIT) compiled languages. A new manual titled >> "OProfile JIT agent developer guide" is provided to aid developers. >> >> Added Xen support for IA64 >> >> Added AVR32 support >> >> Add '--root' option to opreport which act as a replacement (a prefix) >> for the / fs >> >> Performance improvement for 'opreport --xml --details'. >> >> OProfile tends to output obscure error message. Some of these have been fixed. >> >> Changes required to build with gcc 4.3 >> >> Bug fixes >> --------- >> >> Fix sym_offset calculation bug, seen on 64 bit systems with code mapped >> with an address >4G >> >> Fix opcontrol's calculation of kernel address range to include code found >> in sections other than just .text >> >> Fix opcontrol's short forms of --list-events (-l) and --dump (-d) so they >> work for non-root users >> >> Fix MMCR values and counter-to-event mappings on a few 970MP groups >> >> Turn off profiling in hypervisor on 970MP to prevent lost interrupts >> >> Update family10 events and unit_masks files to match the BIOS and Kernel >> Developer's Guide (current, as of May 20, 2008) >> >> Since --xml is not compatible with the --sort option, we warned against >> the use of --sort, but didn't reset the sort options to default. This is >> now fixed. >> >> Fix silent failure of oprof_start for when a counter is missing (e.g, NMI >> watchdog is up) >> >> Change opcontrol to use "-SIG" instead of "-s SIG" since Busybox's >> implementation of "kill" doesn't understand the "-s SIG" option >> >> Fix 'opannotate -s' to work with inlined code >> >> Update POWER6 event files (add new event groupings; make some fixes to others) >> >> Fix "Dangling ESCAPE CODE" error that can occur on Cell BE SPE profiling >> >> Fix opreport for Cell BE SPE profiles to attribute samples to dynamically >> generated call stubs executed from SPE stack (for example, I/O calls to >> the libc library) >> >> Fix loop in 'opcontrol --dump' code when using --session-dir on a network >> drive (clock issues) >> >> Fix problem with 'oparchive --list-files' that can occur if file doesn't exist >> >> Fix user/kernel domain profiling switches for ppc64 architectures >> >> Fix the bfd_get_synthetic_symtab check in m4 macro to work correctly with >> '--with-binutils' configure option >> >> Fix ARM big-endian syscall >> >> bug #1820202, differential profile broken in 0.9.3, is fixed >> >> bug #1717298, mips events have incorrect id numbers, is fixed, all event >> number > 9 was incorrectly setup >> >> bug #1564920, opcontrol does not check if objdump exists is fixed, we error >> out and pinpoint the right error now >> >> bug #1819350, oparchive doesn't work with kernel module, is fixed (need another >> related fix, if you read that I forget about that, ping me please -- phe) >> >> bug #1828566, --xml and -t and callgraph now output symboldata for all >> caller/callees >> >> bug #1930788, opreport error: basic_string::erase >> >> >> Known problems >> -------------- >> >> When using callgraph profiling, it's possible that invalid sample >> files are created (bug #1685267). >> >> Many Alpha ev67 events do not work (bug #931875). >> >> A few Pentium IV events are not supported (bug #841099). >> >> For 2.2 kernels, the module must be compiled as the same user >> that owns the kernel source tree. >> >> With an AMD64 kernel, OProfile must be built in 64 bit mode due to lack >> of kernel support. >> >> opreport -c gives strange output for binaries without symbols. >> >> >> ------------------------------------------------------------------------- >> This SF.net email is sponsored by: Microsoft >> Defy all challenges. Microsoft(R) Visual Studio 2008. >> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ >> _______________________________________________ >> oprofile-list mailing list >> opr...@li... >> https://lists.sourceforge.net/lists/listinfo/oprofile-list >> >> > > > |
From: William C. <wc...@nc...> - 2008-06-03 15:53:06
|
Maynard Johnson wrote: > William Cohen wrote: >> Maynard Johnson wrote: >> >>> We are pleased to announce OProfile 0.9.4 Release Candidate 1. You can download >>> this release at: >>> http://sourceforge.net/project/showfiles.php?group_id=16191 >>> >>> This release has many new features and fixes. The biggest change with this >>> release is the addition of support for profiling JITed code. Java agents that >>> support Java 1.4 (JVMPI) and later (JVMTI) are included with OProfile. See the >>> release notes below for more details. >>> >>> Please test this release candidate and let us know (good or bad) how it works. >>> >>> Thanks. >>> -Maynard Johnson >>> >> Hi Maynard, >> >> Thanks so much for working on a releasing Oprofile 0.9.4. >> >> What is the logic for having oprofile require a special user and group oprofile >> for installation? On rpm builds the build machine might not have the special >> user and groups oprofile this may present an issue. >> > One thing to add to Daniel's reply . . . The configure script puts out a > warning message if the 'oprofile' user and group accounts do not exist, > and 'make install' will fail if they do not exist. Will this be a > problem for your rpm build machines? Please look at the new file in > oprofile's root source directory called README_PACKAGERS for additional > info on spec file changes that are needed to support these required user > and group accounts and let me know if you see any problems with this. >> Noticed that the Fedora 9 kernel (2.6.25.3-18.fc9) on x86_64 started with >> "nmi_watchdog=2" doesn't work with oprofile because performance counter 0 is >> used for the watchdog. There were some patches submitted earlier earlier that >> allowed the performance counter allocation to occur even when some performance >> counter are already reserved. Would it be possible to roll this patch into >> 0.9.4? Below is a link to the related email. >> >> http://marc.info/?l=oprofile-list&m=120716521708920&w=2 I didn't find the README_PACKAGERS file in the oprofile-0.9.4-rc1. It appears to be missing. I took a look to see if there were other discrepancies between the 0.9.4-rc4 and the cvs. Aren't the generated html files generated as part of build? I thought the html were generated as part of the build. Below are files of interest that differd between the 0.9.4-rc1 and cvs. diff -u /tmp/o094rc1.files /tmp/orawhide.files +./doc/buffers.dia -./doc/internals.html +./doc/CodingStyle -./doc/op-jit-devel.html -./doc/oprofile.1 -./doc/oprofile.html +./HACKING -./ltmain.sh -./missing +./README_PACKAGERS RPM packages should be buildable by normal users. rpmbuild makes a separate directory in install the files into, so they can be installed as a normal user in that directory. It might be better to use %defattr in the %file section of the .spec file, so normal users can build the rpm. The real rpm install will give the files the appropriate group/owner. Which files are the ones that should be owned by oprofile? >> > Will, it looks to me like John gave his approval for that patch on April > 16. Assuming that's the case, once you commit it, I can spin a release > candidate 2 if you can test it. I didn't see the approval, I checked the patch in today, so it doesn't hold up the spin. -Will |
From: Maynard J. <may...@us...> - 2008-06-03 22:14:39
|
William Cohen wrote: > Maynard Johnson wrote: >> William Cohen wrote: >>> Maynard Johnson wrote: >>> >>>> We are pleased to announce OProfile 0.9.4 Release Candidate 1. You >>>> can download this release at: >>>> http://sourceforge.net/project/showfiles.php?group_id=16191 >>>> >>>> This release has many new features and fixes. The biggest change >>>> with this release is the addition of support for profiling JITed >>>> code. Java agents that support Java 1.4 (JVMPI) and later (JVMTI) >>>> are included with OProfile. See the release notes below for more >>>> details. >>>> >>>> Please test this release candidate and let us know (good or bad) how >>>> it works. >>>> >>>> Thanks. >>>> -Maynard Johnson >>>> >>> Hi Maynard, >>> >>> Thanks so much for working on a releasing Oprofile 0.9.4. >>> >>> What is the logic for having oprofile require a special user and >>> group oprofile for installation? On rpm builds the build machine >>> might not have the special user and groups oprofile this may present >>> an issue. >>> >> One thing to add to Daniel's reply . . . The configure script puts out a >> warning message if the 'oprofile' user and group accounts do not exist, >> and 'make install' will fail if they do not exist. Will this be a >> problem for your rpm build machines? Please look at the new file in >> oprofile's root source directory called README_PACKAGERS for additional >> info on spec file changes that are needed to support these required user >> and group accounts and let me know if you see any problems with this. >>> Noticed that the Fedora 9 kernel (2.6.25.3-18.fc9) on x86_64 >>> started with "nmi_watchdog=2" doesn't work with oprofile because >>> performance counter 0 is used for the watchdog. There were some >>> patches submitted earlier earlier that allowed the performance >>> counter allocation to occur even when some performance counter are >>> already reserved. Would it be possible to roll this patch into 0.9.4? >>> Below is a link to the related email. >>> >>> http://marc.info/?l=oprofile-list&m=120716521708920&w=2 > > I didn't find the README_PACKAGERS file in the oprofile-0.9.4-rc1. It Oops. Forgot to add this new file to EXTRA_DIST in Makefile.am. Thanks for catching it. > appears to be missing. I took a look to see if there were other > discrepancies between the 0.9.4-rc4 and the cvs. Aren't the generated > html files generated as part of build? I thought the html were generated Looking back at 0.9.3 and 0.9.2, html files were included in the release tar balls. > as part of the build. Below are files of interest that differd between > the 0.9.4-rc1 and cvs. > > diff -u /tmp/o094rc1.files /tmp/orawhide.files > > +./doc/buffers.dia > -./doc/internals.html > +./doc/CodingStyle > -./doc/op-jit-devel.html > -./doc/oprofile.1 > -./doc/oprofile.html > +./HACKING > -./ltmain.sh > -./missing > +./README_PACKAGERS > > RPM packages should be buildable by normal users. rpmbuild makes a > separate directory in install the files into, so they can be installed > as a normal user in that directory. It might be better to use %defattr > in the %file section of the .spec file, so normal users can build the > rpm. The real rpm install will give the files the appropriate > group/owner. Which files are the ones that should be owned by oprofile? There aren't any files distributed with oprofile that are owned by the new 'oprofile' account. This account is used for converting JIT dump files to ELF files instead of having root do the conversion (which it was felt might open a Trojan horse type of security hole). Yes, it would certainly be preferable to allow a normal user to run rpmbuild for oprofile -- and they can -- assuming the special user and group accounts exist on the build machine. But I take it that you would also prefer to build the rpm without having to create the 'oprofile' account on the build machine, right? The README_PACKAGERS file suggests to RPM packagers that they add code in the spec file to create the 'oprofile' account at rpm install time. Is this, along with the configure warning enough to cover the bases? (But maybe not all RPM packagers would read the README_PACKAGERS file.) If we remove the 'make install' error and someone attempts to do JIT profiling without the 'oprofile' account existing, a runtime error would be logged in the oprofiled.log. This is a pretty "quiet" error that not everyone knows how to find, but at least the user would know something didn't work right, since the output would be obviously incorrect. We could beef up the user guide to highlight in more places the need for the 'oprofile account'. Will, I'm not terribly familiar with the rpmbuild process . . . can you think of any other alternatives? Maybe an alternate install target for rpmbuild that doesn't error out if the 'oprofile' account doesn't exist? Anyone else have suggestions or comments? Regards, -Maynard > >>> >> Will, it looks to me like John gave his approval for that patch on April >> 16. Assuming that's the case, once you commit it, I can spin a release >> candidate 2 if you can test it. > > I didn't see the approval, I checked the patch in today, so it doesn't > hold up the spin. > > -Will |
From: John L. <le...@mo...> - 2008-06-04 00:52:34
|
On Tue, Jun 03, 2008 at 05:14:30PM -0500, Maynard Johnson wrote: > > appears to be missing. I took a look to see if there were other > > discrepancies between the 0.9.4-rc4 and the cvs. Aren't the generated > > html files generated as part of build? I thought the html were generated > Looking back at 0.9.3 and 0.9.2, html files were included in the release tar balls. Yes, we don't want to require things like xsltproc just to be able to read the documentation. > a pretty "quiet" error that not everyone knows how to find, but at least the > user would know something didn't work right, since the output would be obviously > incorrect. We could beef up the user guide to highlight in more places the need > for the 'oprofile account'. Presumably we can still *warn* at the very least. regards john |
From: William C. <wc...@nc...> - 2008-06-04 16:22:24
|
Maynard Johnson wrote: <snip> > Yes, it would certainly be preferable to allow a normal user to run > rpmbuild for oprofile -- and they can -- assuming the special user and > group accounts exist on the build machine. But I take it that you would > also prefer to build the rpm without having to create the 'oprofile' > account on the build machine, right? > > The README_PACKAGERS file suggests to RPM packagers that they add code > in the spec file to create the 'oprofile' account at rpm install time. > Is this, along with the configure warning enough to cover the bases? > (But maybe not all RPM packagers would read the README_PACKAGERS file.) > If we remove the 'make install' error and someone attempts to do JIT > profiling without the 'oprofile' account existing, a runtime error would > be logged in the oprofiled.log. This is a pretty "quiet" error that not > everyone knows how to find, but at least the user would know something > didn't work right, since the output would be obviously incorrect. We > could beef up the user guide to highlight in more places the need for > the 'oprofile account'. There are other packages that add users/groups when they are installed. for example oprofile has stapdev and stapusr groups to restrict people how they use the tool. However, these are only used when the systemtap is run, not when systemtap is built. These groups are added with a %pre in the rpm spec file. Should the executable check that the oprofile group/user are there? Shouldn't it be sufficient for the tools at runtime to get an error? Maybe make clear in the error message how to resolve the problem. > > Will, I'm not terribly familiar with the rpmbuild process . . . can you > think of any other alternatives? Maybe an alternate install target for > rpmbuild that doesn't error out if the 'oprofile' account doesn't > exist? Anyone else have suggestions or comments? For the time being I just added a patch that commented out the oprofile group and user check in the make file. I have put the an attempt at packaging oprofile-0.9.4-rc1 at: http://people.redhat.com/wcohen/oprofile-0.9.4rc1-1.fc9.src.rpm Has this been tested with selinux support enabled? -Will > > Regards, > -Maynard > > >> >>>> >>> Will, it looks to me like John gave his approval for that patch on April >>> 16. Assuming that's the case, once you commit it, I can spin a release >>> candidate 2 if you can test it. >> >> I didn't see the approval, I checked the patch in today, so it doesn't >> hold up the spin. >> >> -Will > > |
From: Maynard J. <may...@us...> - 2008-06-05 00:56:01
|
William Cohen wrote: > Maynard Johnson wrote: > > <snip> > > >> Yes, it would certainly be preferable to allow a normal user to run >> rpmbuild for oprofile -- and they can -- assuming the special user and >> group accounts exist on the build machine. But I take it that you would >> also prefer to build the rpm without having to create the 'oprofile' >> account on the build machine, right? >> >> The README_PACKAGERS file suggests to RPM packagers that they add code >> in the spec file to create the 'oprofile' account at rpm install time. >> Is this, along with the configure warning enough to cover the bases? >> (But maybe not all RPM packagers would read the README_PACKAGERS file.) >> If we remove the 'make install' error and someone attempts to do JIT >> profiling without the 'oprofile' account existing, a runtime error would >> be logged in the oprofiled.log. This is a pretty "quiet" error that not >> everyone knows how to find, but at least the user would know something >> didn't work right, since the output would be obviously incorrect. We >> could beef up the user guide to highlight in more places the need for >> the 'oprofile account'. >> > > There are other packages that add users/groups when they are installed. for > example oprofile has stapdev and stapusr groups to restrict people how they use > the tool. However, these are only used when the systemtap is run, not when > systemtap is built. These groups are added with a %pre in the rpm spec file. > Should the executable check that the oprofile group/user are there? Shouldn't it > be sufficient for the tools at runtime to get an error? Maybe make clear in the > error message how to resolve the problem. > > >> Will, I'm not terribly familiar with the rpmbuild process . . . can you >> think of any other alternatives? Maybe an alternate install target for >> rpmbuild that doesn't error out if the 'oprofile' account doesn't >> exist? Anyone else have suggestions or comments? >> > > For the time being I just added a patch that commented out the oprofile group > and user check in the make file. I have put the an attempt at packaging > oprofile-0.9.4-rc1 at: > > http://people.redhat.com/wcohen/oprofile-0.9.4rc1-1.fc9.src.rpm > > Has this been tested with selinux support enabled? > Not that I know of. Can you test that, Will? I'm of the opinion that we should remove the error in 'make install' and just put out a warning message if the 'oprofile' user/group does not exist. We shouldn't force people to create this account if they don't need or want to profile JITed code. But I need to come up with a technique to ensure the message isn't lost in the clutter of the rest of the install messages. I'm going to be out for a few days without internet access, so I'll work on fixing this up and put out a new release candidate when I return next week (Tuesday, at the earliest). -Maynard > -Will > > > >> Regards, >> -Maynard >> >> >> >>>>> >>>>> >>>> Will, it looks to me like John gave his approval for that patch on April >>>> 16. Assuming that's the case, once you commit it, I can spin a release >>>> candidate 2 if you can test it. >>>> >>> I didn't see the approval, I checked the patch in today, so it doesn't >>> hold up the spin. >>> >>> -Will >>> >> > > > |