[Buildroot] [PATCH 1/6] linux/perf: really do not build the documentation

[Buildroot] [PATCH 1/6] linux/perf: really do not build the documentation

[Yann E. MORIN]

The perf buildsystem, inside the kernel, is not really amenable to be
easily used...

Regarding the documentation, it will forcefully try to detect asciidoc
and, with the latest versions, xmlto, completely disregarding what the
user may provide.

We currently pass ASCIIDOC= (the empty string) on the make command line,
as an attempt to disable building the documentation, but that has no
effect whatsoever on perf: that variable is not passed down to the
sub-sub-make (yes, a two-level depth) that is responsible for building
the documentation.

We really do not want to build any of the documentation (the user can
refer to the documentation on his own development machine), so we use a
little dirty trick: we provide a GNUmakefile beside the existing
Makefile for the documentation; GNUmakefile always takes precedence over
a Makefile when both are present. We only provide a catch-all-no-recipe
rule in that GNUmakefile, so it really does nothing useful, except avoid
building the documentation.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas De Schampheleire <patrickdepinguin@gmail.com>
---
 linux/linux-tool-perf.mk | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/linux/linux-tool-perf.mk b/linux/linux-tool-perf.mk
index 1070a1c..f9b295f 100644
--- a/linux/linux-tool-perf.mk
+++ b/linux/linux-tool-perf.mk
@@ -24,8 +24,7 @@ PERF_MAKE_FLAGS = \
 	NO_LIBPYTHON=1 \
 	DESTDIR=$(TARGET_DIR) \
 	prefix=/usr \
-	WERROR=0 \
-	ASCIIDOC=
+	WERROR=0
 
 # We need to pass an argument to ld for setting the endianness when
 # building it for MIPS architecture, otherwise the default one will
@@ -68,6 +67,19 @@ else
 PERF_MAKE_FLAGS += NO_LIBELF=1 NO_DWARF=1
 endif
 
+# We really do not want to build the perf documentation, because it
+# has stringent requirement on the documentation generation tools,
+# like xmlto and asciidoc), which may be lagging behind on some
+# distributions.
+# We name it 'GNUmakefile' so that GNU make will use it instead of
+# the existing 'Makefile'.
+define PERF_DISABLE_DOCUMENTATION
+	if [ -f $(@D)/tools/perf/Documentation/Makefile ]; then \
+		printf "%%:\n\t@:\n" >$(@D)/tools/perf/Documentation/GNUmakefile; \
+	fi
+endef
+LINUX_POST_PATCH_HOOKS += PERF_DISABLE_DOCUMENTATION
+
 # O must be redefined here to overwrite the one used by Buildroot for
 # out of tree build. We build perf in $(@D)/tools/perf/ and not just
 # $(@D) so that it isn't built in the root directory of the kernel
-- 
1.9.1

[Buildroot] [PATCH 2/6] linux/perf: honour the number of parallel jobs

[Yann E. MORIN]

perf does not honour the -j flags we pass to make; it yet again tries to
reinvent the wheel and by default uses the number of CPUs as the number
of parallel jobs.

Fortunately, in their inifinite wisdom, the insane developpers of the
perf buildsystem were kind enough to provide us with a variable we can
set to specify the number of parallel jobs.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas De Schampheleire <patrickdepinguin@gmail.com>
---
 linux/linux-tool-perf.mk | 1 +
 1 file changed, 1 insertion(+)

diff --git a/linux/linux-tool-perf.mk b/linux/linux-tool-perf.mk
index f9b295f..a5fbf21 100644
--- a/linux/linux-tool-perf.mk
+++ b/linux/linux-tool-perf.mk
@@ -16,6 +16,7 @@ endif
 
 PERF_MAKE_FLAGS = \
 	$(LINUX_MAKE_FLAGS) \
+	JOBS=$(PARALLEL_JOBS) \
 	ARCH=$(PERF_ARCH) \
 	NO_LIBAUDIT=1 \
 	NO_NEWT=1 \
-- 
1.9.1

[Buildroot] [PATCH 0/6] linux: build host perf (branch yem/linux-host-perf)

[Yann E. MORIN]

Hello All!

This short series make it possible to build a host perf.

When run on the target, perf may generate raw dumps. Those dumps are to
be parsed by perf itself to gfenerate proper reports; this can (and
should probably) be done off-line ona fast machine, rather than on the
target.

First, we apply a number of fixes to the perf linux tool:
  - really do not build the documentation
  - honour the number of parallel jobs
  - really disable unavailable features
  - conditioanlly add support for gz and xz

Then, we build perf for the target in an empty directory, rather than
in-place, so that we do not mix the target .o files with the host .o
files.

Finally, we do build the host perf, alsao in its own directory.

Regards,
Yann E. MORIN.


The following changes since commit 0b7b84310c3781c346eea4c114c8cb4e97bbd8ea:

  olimex_imx233_olinuxino_defconfig: genimage support (2016-03-11 13:14:54 +0100)

are available in the git repository at:

  git://git.busybox.net/~ymorin/git/buildroot yem/linux-host-perf

for you to fetch changes up to b5f50fa3c517ddd37dbf339c011f6bf0621f55ac:

  linux/perf: build the host perf tool (2016-03-11 18:11:58 +0100)

----------------------------------------------------------------
Yann E. MORIN (6):
      linux/perf: really do not build the documentation
      linux/perf: honour the number of parallel jobs
      linux/perf: forcefully disable the features with missing dependencies
      linux/perf: conditionally enable support for gz/xz compression
      linux/perf: build in a sub-directory
      linux/perf: build the host perf tool

 linux/Config.tools.in    |  4 +++
 linux/linux-tool-perf.mk | 93 ++++++++++++++++++++++++++++++++++++++++++------
 2 files changed, 87 insertions(+), 10 deletions(-)

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'

[Buildroot] [PATCH 3/6] linux/perf: forcefully disable the features with missing dependencies

[Yann E. MORIN]

Forcefully disable the features that have optional dependencies that are
not enabled in Buildroot.

Disable support for bionic since, well, we're not Android.

Slightly re-order the variables to have semantically-related variables
together, with features last.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas De Schampheleire <patrickdepinguin@gmail.com>
---
 linux/linux-tool-perf.mk | 13 ++++++++++---
 1 file changed, 10 insertions(+), 3 deletions(-)

diff --git a/linux/linux-tool-perf.mk b/linux/linux-tool-perf.mk
index a5fbf21..d4b8f0e 100644
--- a/linux/linux-tool-perf.mk
+++ b/linux/linux-tool-perf.mk
@@ -18,14 +18,15 @@ PERF_MAKE_FLAGS = \
 	$(LINUX_MAKE_FLAGS) \
 	JOBS=$(PARALLEL_JOBS) \
 	ARCH=$(PERF_ARCH) \
+	DESTDIR=$(TARGET_DIR) \
+	prefix=/usr \
+	WERROR=0 \
 	NO_LIBAUDIT=1 \
 	NO_NEWT=1 \
 	NO_GTK2=1 \
 	NO_LIBPERL=1 \
 	NO_LIBPYTHON=1 \
-	DESTDIR=$(TARGET_DIR) \
-	prefix=/usr \
-	WERROR=0
+	NO_LIBBIONIC=1
 
 # We need to pass an argument to ld for setting the endianness when
 # building it for MIPS architecture, otherwise the default one will
@@ -52,14 +53,20 @@ endif
 
 ifeq ($(BR2_PACKAGE_SLANG),y)
 PERF_DEPENDENCIES += slang
+else
+PERF_MAKE_FLAGS += NO_SLANG=1
 endif
 
 ifeq ($(BR2_PACKAGE_LIBUNWIND),y)
 PERF_DEPENDENCIES += libunwind
+else
+PERF_MAKE_FLAGS += NO_LIBUNWIND=1
 endif
 
 ifeq ($(BR2_PACKAGE_NUMACTL),y)
 PERF_DEPENDENCIES += numactl
+else
+PERF_MAKE_FLAGS += NO_LIBNUMA=1
 endif
 
 ifeq ($(BR2_PACKAGE_ELFUTILS),y)
-- 
1.9.1

[Buildroot] [PATCH 4/6] linux/perf: conditionally enable support for gz/xz compression

[Yann E. MORIN]

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas De Schampheleire <patrickdepinguin@gmail.com>
---
 linux/linux-tool-perf.mk | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/linux/linux-tool-perf.mk b/linux/linux-tool-perf.mk
index d4b8f0e..8143474 100644
--- a/linux/linux-tool-perf.mk
+++ b/linux/linux-tool-perf.mk
@@ -75,6 +75,19 @@ else
 PERF_MAKE_FLAGS += NO_LIBELF=1 NO_DWARF=1
 endif
 
+ifeq ($(BR2_PACKAGE_ZLIB),y)
+PERF_DEPENDENCIES += zlib
+else
+PERF_MAKE_FLAGS += NO_ZLIB=1
+endif
+
+# lzma is provided by xz
+ifeq ($(BR2_PACKAGE_XZ),y)
+PERF_DEPENDENCIES += xz
+else
+PERF_MAKE_FLAGS += NO_LZMA=1
+endif
+
 # We really do not want to build the perf documentation, because it
 # has stringent requirement on the documentation generation tools,
 # like xmlto and asciidoc), which may be lagging behind on some
-- 
1.9.1

[Buildroot] [PATCH 5/6] linux/perf: build in a sub-directory

[Yann E. MORIN]

Currently, we only build perf for the target.

However, analysing some of the results gathered by perf (aka trace) may
require running perf (aka trace) on the host, so we'll have to build
perf for the host too.

To avoid mixing target and host binaries, we need to build them in
separate directories.

This commit only moves the target build, in preparation of the host
build to come later.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas De Schampheleire <patrickdepinguin@gmail.com>
---
 linux/linux-tool-perf.mk | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/linux/linux-tool-perf.mk b/linux/linux-tool-perf.mk
index 8143474..f584955 100644
--- a/linux/linux-tool-perf.mk
+++ b/linux/linux-tool-perf.mk
@@ -120,15 +120,16 @@ define PERF_BUILD_CMDS
 			fi \
 		fi \
 	fi
+	$(Q)mkdir -p $(@D)/tools/perf/br-target
 	$(TARGET_MAKE_ENV) $(MAKE1) $(PERF_MAKE_FLAGS) \
-		-C $(@D)/tools/perf O=$(@D)/tools/perf/
+		-C $(@D)/tools/perf O=$(@D)/tools/perf/br-target
 endef
 
 # After installation, we remove the Perl and Python scripts from the
 # target.
 define PERF_INSTALL_TARGET_CMDS
 	$(TARGET_MAKE_ENV) $(MAKE1) $(PERF_MAKE_FLAGS) \
-		-C $(@D)/tools/perf O=$(@D)/tools/perf/ install
+		-C $(@D)/tools/perf O=$(@D)/tools/perf/br-target install
 	$(RM) -rf $(TARGET_DIR)/usr/libexec/perf-core/scripts/
 	$(RM) -rf $(TARGET_DIR)/usr/libexec/perf-core/tests/
 endef
-- 
1.9.1

[Buildroot] [PATCH 6/6] linux/perf: build the host perf tool

[Yann E. MORIN]

Currently, we only build the target variant of the perf tool. However,
perf on the target may generate a bunch of data files that may have to
be analysed on the host.

There is no host-variant of the linux-tools infrastructure, so we just
build the host perf at the same time we build the target one.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas De Schampheleire <patrickdepinguin@gmail.com>
---
 linux/Config.tools.in    |  4 ++++
 linux/linux-tool-perf.mk | 51 ++++++++++++++++++++++++++++++++++++++++++------
 2 files changed, 49 insertions(+), 6 deletions(-)

diff --git a/linux/Config.tools.in b/linux/Config.tools.in
index 24ef8cd..a01b53e 100644
--- a/linux/Config.tools.in
+++ b/linux/Config.tools.in
@@ -26,4 +26,8 @@ config BR2_LINUX_KERNEL_TOOL_PERF
 
 	  https://perf.wiki.kernel.org/
 
+config BR2_LINUX_KERNEL_TOOL_HOST_PERF
+    bool "host-perf"
+    depends on BR2_LINUX_KERNEL_TOOL_PERF
+
 endmenu
diff --git a/linux/linux-tool-perf.mk b/linux/linux-tool-perf.mk
index f584955..c352677 100644
--- a/linux/linux-tool-perf.mk
+++ b/linux/linux-tool-perf.mk
@@ -41,6 +41,26 @@ else ifeq ($(BR2_mipsel)$(BR2_mips64el),y)
 PERF_MAKE_FLAGS += LD="$(TARGET_LD) -EL"
 endif
 
+# For the host, we inherit the same flags as for the target,
+# overriding just those that are needed.
+HOST_PERF_MAKE_FLAGS = \
+	$(PERF_MAKE_FLAGS) \
+	CFLAGS="$(HOST_CFLAGS) -I$(HOST_DIR)/usr/include/elfutils" \
+	LDFLAGS="$(HOST_LDFLAGS)" \
+	LD="$(HOSTLD)" \
+	ARCH=$(PERF_ARCH) \
+	CROSS_COMPILE= \
+	DESTDIR=$(HOST_DIR)
+
+# No host-packages in Buildroot, but we don't let it depend on the
+# libraries provided by the host system (they may be there or not,
+# depending on the user's system, but we won't play that game).
+HOST_PERF_MAKE_FLAGS += \
+	NO_SLANG=1 \
+	NO_LIBUNWIND=1 \
+	NO_LIBNUMA=1 \
+	NO_LIBELF=1 NO_DWARF=1
+
 # The call to backtrace() function fails for ARC, because for some
 # reason the unwinder from libgcc returns early. Thus the usage of
 # backtrace() should be disabled in perf explicitly: at build time
@@ -49,6 +69,7 @@ endif
 # instead of the complete backtrace.
 ifeq ($(BR2_arc),y)
 PERF_MAKE_FLAGS += NO_BACKTRACE=1
+HOST_PERF_MAKE_FLAGS += NO_BACKTRACE=1
 endif
 
 ifeq ($(BR2_PACKAGE_SLANG),y)
@@ -76,14 +97,14 @@ PERF_MAKE_FLAGS += NO_LIBELF=1 NO_DWARF=1
 endif
 
 ifeq ($(BR2_PACKAGE_ZLIB),y)
-PERF_DEPENDENCIES += zlib
+PERF_DEPENDENCIES += zlib host-zlib
 else
 PERF_MAKE_FLAGS += NO_ZLIB=1
 endif
 
 # lzma is provided by xz
 ifeq ($(BR2_PACKAGE_XZ),y)
-PERF_DEPENDENCIES += xz
+PERF_DEPENDENCIES += xz host-xz
 else
 PERF_MAKE_FLAGS += NO_LZMA=1
 endif
@@ -101,10 +122,26 @@ define PERF_DISABLE_DOCUMENTATION
 endef
 LINUX_POST_PATCH_HOOKS += PERF_DISABLE_DOCUMENTATION
 
-# O must be redefined here to overwrite the one used by Buildroot for
-# out of tree build. We build perf in $(@D)/tools/perf/ and not just
-# $(@D) so that it isn't built in the root directory of the kernel
-# sources.
+# O must be redefined in the commands to overwrite the one used by
+# Buildroot for out of tree build. This is valid for both the host
+# and the target builds.
+#
+# We specify the directory where to build perf so that:
+#  - we don't mix the host and the target variants,
+#  - it isn't built in the root directory of the kernel sources.
+
+ifeq ($(BR2_LINUX_KERNEL_TOOL_HOST_PERF),y)
+define HOST_PERF_BUILD_CMDS
+	$(Q)mkdir -p $(@D)/tools/perf/br-host
+	$(HOST_MAKE_ENV) $(MAKE1) $(HOST_PERF_MAKE_FLAGS) \
+		-C $(@D)/tools/perf O=$(@D)/tools/perf/br-host
+endef
+define HOST_PERF_INSTALL_CMDS
+	$(HOST_MAKE_ENV) $(MAKE1) $(HOST_PERF_MAKE_FLAGS) \
+		-C $(@D)/tools/perf O=$(@D)/tools/perf/br-host install
+endef
+endif
+
 define PERF_BUILD_CMDS
 	$(Q)if test ! -f $(@D)/tools/perf/Makefile ; then \
 		echo "Your kernel version is too old and does not have the perf tool." ; \
@@ -123,6 +160,7 @@ define PERF_BUILD_CMDS
 	$(Q)mkdir -p $(@D)/tools/perf/br-target
 	$(TARGET_MAKE_ENV) $(MAKE1) $(PERF_MAKE_FLAGS) \
 		-C $(@D)/tools/perf O=$(@D)/tools/perf/br-target
+	$(HOST_PERF_BUILD_CMDS)
 endef
 
 # After installation, we remove the Perl and Python scripts from the
@@ -132,4 +170,5 @@ define PERF_INSTALL_TARGET_CMDS
 		-C $(@D)/tools/perf O=$(@D)/tools/perf/br-target install
 	$(RM) -rf $(TARGET_DIR)/usr/libexec/perf-core/scripts/
 	$(RM) -rf $(TARGET_DIR)/usr/libexec/perf-core/tests/
+	$(HOST_PERF_INSTALL_CMDS)
 endef
-- 
1.9.1

[Buildroot] [PATCH 0/6] linux: build host perf (branch yem/linux-host-perf)

[Thomas Petazzoni]

Hello,

On Fri, 11 Mar 2016 19:19:56 +0100, Yann E. MORIN wrote:

>       linux/perf: really do not build the documentation
>       linux/perf: honour the number of parallel jobs
>       linux/perf: forcefully disable the features with missing dependencies
>       linux/perf: conditionally enable support for gz/xz compression

I've applied those four first patches.

>       linux/perf: build in a sub-directory

This one looks OK to me, but it doesn't make much sense applying it
without applying the last patch, on which I have comments.

>       linux/perf: build the host perf tool

I'll reply on this one.

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com

[Buildroot] [PATCH 6/6] linux/perf: build the host perf tool

[Thomas Petazzoni]

Yann,

On Fri, 11 Mar 2016 19:19:59 +0100, Yann E. MORIN wrote:
> Currently, we only build the target variant of the perf tool. However,
> perf on the target may generate a bunch of data files that may have to
> be analysed on the host.
> 
> There is no host-variant of the linux-tools infrastructure, so we just
> build the host perf at the same time we build the target one.
> 
> Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
> Cc: Thomas De Schampheleire <patrickdepinguin@gmail.com>

First of all, I have to say I hate this target package that builds and
installs host stuff. But, since I don't have a better solution to
offer, and it's anyway not the only package in this case (qt/qt5 are
also target packages, and they install stuff in host), I'm fine with
the general approach.

I have some comments about the details, though.

> ---
>  linux/Config.tools.in    |  4 ++++
>  linux/linux-tool-perf.mk | 51 ++++++++++++++++++++++++++++++++++++++++++------
>  2 files changed, 49 insertions(+), 6 deletions(-)
> 
> diff --git a/linux/Config.tools.in b/linux/Config.tools.in
> index 24ef8cd..a01b53e 100644
> --- a/linux/Config.tools.in
> +++ b/linux/Config.tools.in
> @@ -26,4 +26,8 @@ config BR2_LINUX_KERNEL_TOOL_PERF
>  
>  	  https://perf.wiki.kernel.org/
>  
> +config BR2_LINUX_KERNEL_TOOL_HOST_PERF
> +    bool "host-perf"
> +    depends on BR2_LINUX_KERNEL_TOOL_PERF

Indentation should just tab.

> +# For the host, we inherit the same flags as for the target,
> +# overriding just those that are needed.

I don't really understand this, it seems really weird to me, and I
would find it much more future proof to have separate target and host
flags. Otherwise, any addition of a flag to the target variant can
break in subtle way the host variant.

> +HOST_PERF_MAKE_FLAGS = \
> +	$(PERF_MAKE_FLAGS) \
> +	CFLAGS="$(HOST_CFLAGS) -I$(HOST_DIR)/usr/include/elfutils" \

We don't have this -I for the target variant. Why do we need it for the
host variant?

> +	LDFLAGS="$(HOST_LDFLAGS)" \
> +	LD="$(HOSTLD)" \
> +	ARCH=$(PERF_ARCH) \
> +	CROSS_COMPILE= \
> +	DESTDIR=$(HOST_DIR)
> +
> +# No host-packages in Buildroot, but we don't let it depend on the
> +# libraries provided by the host system (they may be there or not,
> +# depending on the user's system, but we won't play that game).

This sentence is not really clear IMO. What about:

# For the following optional dependencies of perf, Buildroot doesn't
# provide any host package, and we don't want to rely on those
# libraries being provided by the host system.

> +HOST_PERF_MAKE_FLAGS += \
> +	NO_SLANG=1 \
> +	NO_LIBUNWIND=1 \
> +	NO_LIBNUMA=1 \
> +	NO_LIBELF=1 NO_DWARF=1
> +
>  # The call to backtrace() function fails for ARC, because for some
>  # reason the unwinder from libgcc returns early. Thus the usage of
>  # backtrace() should be disabled in perf explicitly: at build time
> @@ -49,6 +69,7 @@ endif
>  # instead of the complete backtrace.
>  ifeq ($(BR2_arc),y)
>  PERF_MAKE_FLAGS += NO_BACKTRACE=1
> +HOST_PERF_MAKE_FLAGS += NO_BACKTRACE=1

Why ?

>  endif
>  
>  ifeq ($(BR2_PACKAGE_SLANG),y)
> @@ -76,14 +97,14 @@ PERF_MAKE_FLAGS += NO_LIBELF=1 NO_DWARF=1
>  endif
>  
>  ifeq ($(BR2_PACKAGE_ZLIB),y)
> -PERF_DEPENDENCIES += zlib
> +PERF_DEPENDENCIES += zlib host-zlib

This is the part I'm really not a big fan of. You enable a target
package, and it affects the configuration of a host package. Not nice.

Is there an actual connection between zlib/xz support in the target
perf and the need to have it in host-perf? Like is zlib/xz support used
to compress the data output by the target perf, which needs to be
uncompressed by the host perf ?

Thanks!

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com

[Buildroot] [PATCH 6/6] linux/perf: build the host perf tool

[Yann E. MORIN]

Thomas, All,

On 2016-03-19 14:55 +0100, Thomas Petazzoni spake thusly:
> On Fri, 11 Mar 2016 19:19:59 +0100, Yann E. MORIN wrote:
> > Currently, we only build the target variant of the perf tool. However,
> > perf on the target may generate a bunch of data files that may have to
> > be analysed on the host.
> > 
> > There is no host-variant of the linux-tools infrastructure, so we just
> > build the host perf at the same time we build the target one.
> > 
> > Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
> > Cc: Thomas De Schampheleire <patrickdepinguin@gmail.com>
> 
> First of all, I have to say I hate this target package that builds and
> installs host stuff.

Yeah, that bothered me too, enough that I had to explain it in the
commit log.

[--SNIP--]
> > diff --git a/linux/Config.tools.in b/linux/Config.tools.in
> > index 24ef8cd..a01b53e 100644
> > --- a/linux/Config.tools.in
> > +++ b/linux/Config.tools.in
> > @@ -26,4 +26,8 @@ config BR2_LINUX_KERNEL_TOOL_PERF
> >  
> >  	  https://perf.wiki.kernel.org/
> >  
> > +config BR2_LINUX_KERNEL_TOOL_HOST_PERF
> > +    bool "host-perf"
> > +    depends on BR2_LINUX_KERNEL_TOOL_PERF
> Indentation should just tab.

Damned... :-)

> > +# For the host, we inherit the same flags as for the target,
> > +# overriding just those that are needed.
> 
> I don't really understand this, it seems really weird to me, and I
> would find it much more future proof to have separate target and host
> flags. Otherwise, any addition of a flag to the target variant can
> break in subtle way the host variant.

Well, I'm not a perf expert, but my reasoning is that, if the target
perf stores info about, say, function unwinding, the host perf has to
be able to parse it. Conversely, if the target perf does not support
tracing function unwinding, there is no point in having that support
in the host perf.

So, we just use the same set of options, except for the compiler suite
and the destination directory.

> > +HOST_PERF_MAKE_FLAGS = \
> > +	$(PERF_MAKE_FLAGS) \
> > +	CFLAGS="$(HOST_CFLAGS) -I$(HOST_DIR)/usr/include/elfutils" \
> 
> We don't have this -I for the target variant. Why do we need it for the
> host variant?

I can retry without it, but IIRC it did not work. I'll double-check.

> > +	LDFLAGS="$(HOST_LDFLAGS)" \
> > +	LD="$(HOSTLD)" \
> > +	ARCH=$(PERF_ARCH) \
> > +	CROSS_COMPILE= \
> > +	DESTDIR=$(HOST_DIR)
> > +
> > +# No host-packages in Buildroot, but we don't let it depend on the
> > +# libraries provided by the host system (they may be there or not,
> > +# depending on the user's system, but we won't play that game).
> 
> This sentence is not really clear IMO. What about:
> 
> # For the following optional dependencies of perf, Buildroot doesn't
> # provide any host package, and we don't want to rely on those
> # libraries being provided by the host system.

Yeah, better. Thanks!

> > +HOST_PERF_MAKE_FLAGS += \
> > +	NO_SLANG=1 \
> > +	NO_LIBUNWIND=1 \
> > +	NO_LIBNUMA=1 \
> > +	NO_LIBELF=1 NO_DWARF=1
> > +
> >  # The call to backtrace() function fails for ARC, because for some
> >  # reason the unwinder from libgcc returns early. Thus the usage of
> >  # backtrace() should be disabled in perf explicitly: at build time
> > @@ -49,6 +69,7 @@ endif
> >  # instead of the complete backtrace.
> >  ifeq ($(BR2_arc),y)
> >  PERF_MAKE_FLAGS += NO_BACKTRACE=1
> > +HOST_PERF_MAKE_FLAGS += NO_BACKTRACE=1
> 
> Why ?

Ditto the explanation about inheriting the target options, above.

> >  endif
> >  
> >  ifeq ($(BR2_PACKAGE_SLANG),y)
> > @@ -76,14 +97,14 @@ PERF_MAKE_FLAGS += NO_LIBELF=1 NO_DWARF=1
> >  endif
> >  
> >  ifeq ($(BR2_PACKAGE_ZLIB),y)
> > -PERF_DEPENDENCIES += zlib
> > +PERF_DEPENDENCIES += zlib host-zlib
> 
> This is the part I'm really not a big fan of. You enable a target
> package, and it affects the configuration of a host package. Not nice.

Ditto, to match what the target perf supports.

However, sicne there is no host variant, there's not much we can do to
clean this mess. And I doubt we want to add a host-linux package just
for host-perf.

> Is there an actual connection between zlib/xz support in the target
> perf and the need to have it in host-perf? Like is zlib/xz support used
> to compress the data output by the target perf, which needs to be
> uncompressed by the host perf ?

TBH, I did not try. I was secretly hoping that Thomas DS would test-run
it... (Niark!)

Regards,
Yann E. MORIN.

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'

[Buildroot] [PATCH 6/6] linux/perf: build the host perf tool

[Arnout Vandecappelle]

On 03/19/16 14:55, Thomas Petazzoni wrote:
> On Fri, 11 Mar 2016 19:19:59 +0100, Yann E. MORIN wrote:
>> >Currently, we only build the target variant of the perf tool. However,
>> >perf on the target may generate a bunch of data files that may have to
>> >be analysed on the host.
>> >
>> >There is no host-variant of the linux-tools infrastructure, so we just
>> >build the host perf at the same time we build the target one.
>> >
>> >Signed-off-by: "Yann E. MORIN"<yann.morin.1998@free.fr>
>> >Cc: Thomas De Schampheleire<patrickdepinguin@gmail.com>
> First of all, I have to say I hate this target package that builds and
> installs host stuff. But, since I don't have a better solution to
> offer, and it's anyway not the only package in this case (qt/qt5 are
> also target packages, and they install stuff in host), I'm fine with
> the general approach.

  I think it's OK to build and install host tools that are bound tightly with 
the target package. E.g. qmake only makes sense when you have qt for the target. 
perf is in the same boat.

  Regards,
  Arnout

-- 
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF