[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <87bnyy6m5s.fsf@sejong.aot.lge.com>
Date: Mon, 27 Jan 2014 14:14:55 +0900
From: Namhyung Kim <namhyung@...nel.org>
To: Jiri Olsa <jolsa@...hat.com>
Cc: Josh Boyer <jwboyer@...oraproject.org>,
Ingo Molnar <mingo@...e.hu>,
Steven Rostedt <rostedt@...dmis.org>,
Arnaldo Carvalho de Melo <acme@...hat.com>,
"Linux-Kernel\@Vger. Kernel. Org" <linux-kernel@...r.kernel.org>
Subject: Re: Weird plugin paths in perf and perf.so binaries with 3.14 merge window
Hi Jiri and Josh,
On Wed, 22 Jan 2014 12:36:30 +0100, Jiri Olsa wrote:
> On Tue, Jan 21, 2014 at 03:02:43PM -0500, Josh Boyer wrote:
>> Hi All,
>>
>> I went to build Linux v3.13-737-g7fe67a1 in Fedora this morning and it
>> resulted in RPM complaining that the perf and perf.so binaries had
>> strings in them that matched the RPM_BUILD_ROOT string. That fails
>> the RPM build.
>>
>> Looking over the logs, I see that the traceevent plugins are getting a
>> rather weird -DPLUGIN_DIR define passed to them. E.g.:
>>
>> gcc -c -g -Wall -I. -I
>> /home/jwboyer/kernel/kernel-3.13.fc21/linux-3.14.0-0.rc0.git1.1.fc21.x86_64/tools/lib/traceevent/../../include
>> '-DPLUGIN_DIR="/home/jwboyer/rpmbuild/BUILDROOT/kernel-3.14.0-0.rc0.git1.1.fc21.x86_64//usr//usr/lib64/traceevent/plugins"'
>> -D_GNU_SOURCE -std=gnu99 -fPIC
>> /home/jwboyer/kernel/kernel-3.13.fc21/linux-3.14.0-0.rc0.git1.1.fc21.x86_64/tools/lib/traceevent/parse-filter.c
>> -o parse-filter.o
>>
>> We're building perf like so:
>>
>> make -s -j8 -C tools/perf V=1 WERROR=0 NO_LIBUNWIND=1
>> HAVE_CPLUS_DEMANGLE=1 NO_GTK2=1 NO_LIBNUMA=1 NO_STRLCPY=1 NO_BIONIC=1
>> prefix=/usr DESTDIR=/home/jwboyer/rpmbuild/BUILDROOT/kernel-3.14.0-0.rc0.git1.1.fc21.x86_64
>> all
>>
>> and installing it via:
>>
>> make -s -j8 -C tools/perf V=1 WERROR=0 NO_LIBUNWIND=1
>> HAVE_CPLUS_DEMANGLE=1 NO_GTK2=1 NO_LIBNUMA=1 NO_STRLCPY=1 NO_BIONIC=1
>> prefix=/usr DESTDIR=/home/jwboyer/rpmbuild/BUILDROOT/kernel-3.14.0-0.rc0.git1.1.fc21.x86_64
>> install-bin
>>
>> make -s -j8 -C tools/perf V=1 WERROR=0 NO_LIBUNWIND=1
>> HAVE_CPLUS_DEMANGLE=1 NO_GTK2=1 NO_LIBNUMA=1 NO_STRLCPY=1 NO_BIONIC=1
>> prefix=/usr DESTDIR=/home/jwboyer/rpmbuild/BUILDROOT/kernel-3.14.0-0.rc0.git1.1.fc21.x86_64
>> install-python_ext
>>
>> This has worked for all the builds up until this point.
>>
>> Somehow the perf and perf.so binaries are getting the string being
>> passed via -DPLUGIN_DIR into them, likely through the libtraceevent.a
>> link. I'm pretty sure that (1) the string being passed is totally
>> broken and should be "/usr/lib64/traceevent/plugins" and (2) that I
>> haven't come close to deciphering how to fix this.
>>
>> So, could you please look this over and see where the define is
>> getting messed up?
I found three things on the issue:
1. perf defines plugin_dir if DESTDIR is set like below:
ifdef DESTDIR
plugindir=$(libdir)/traceevent/plugins
plugindir_SQ= $(subst ','\'',$(prefix)/$(plugindir))
endif
But this is wrong since $plugindir includes $libdir and $libdir
includes $prefix in it. So there'll be two $prefix's in
$plugindir_SQ as you can see above (...//usr//usr/...).
2. And libtraceevent prepends DESTDIR to the plugindir:
ifeq ($(set_plugin_dir),1)
PLUGIN_DIR = -DPLUGIN_DIR="$(DESTDIR)/$(plugin_dir)"
PLUGIN_DIR_SQ = '$(subst ','\'',$(PLUGIN_DIR))'
endif
I don't know if it's desired but as Josh reported it seems not.
Maybe we'd better adding just $prefix part not the $DESTDIR.
3. libtraceevent's default plugin_dir is set like:
ifeq ($(prefix),$(HOME))
override plugin_dir = $(HOME)/.traceevent/plugins
set_plugin_dir := 0
else
override plugin_dir = $(prefix)/lib/traceevent/plugins
endif
But it'd better using lib64/ instead of lib/ on 64 bit machines. We
can reuse same logic (libdir) in the perf Makefile but it's not a big
deal as we can set it on the cmdline so I won't insist strongly. ;-)
Thanks,
Namhyung
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists