The build system is inserting an extra pipe "|" where it shouldn't be, and this is causing a problem. Reproducible: Always # emerge --info Portage 2.2.0_alpha142 (default/linux/amd64/10.0/no-multilib, gcc-4.7.2, glibc-2.16.0, 3.7.0-rc6 x86_64) ================================================================= System uname: Linux-3.7.0-rc6-x86_64-AMD_Phenom-tm-_II_X4_965_Processor-with-gentoo-2.2 Timestamp of tree: Thu, 15 Nov 2012 22:45:01 +0000 ld GNU ld (Linux/GNU Binutils) 2.23.51.0.5.20121110 distcc 3.2rc1 x86_64-pc-linux-gnu [disabled] app-shells/bash: 4.2_p39 dev-lang/python: 2.7.3-r2, 3.2.3-r1 dev-util/cmake: 2.8.9-r1 dev-util/pkgconfig: 0.27.1 sys-apps/baselayout: 2.2 sys-apps/openrc: 9999 sys-apps/sandbox: 2.6 sys-devel/autoconf: 2.13, 2.69 sys-devel/automake: 1.11.6, 1.12.4 sys-devel/binutils: 2.23.51.0.5 sys-devel/gcc: 4.7.2 sys-devel/gcc-config: 1.7.3 sys-devel/libtool: 2.4.2 sys-devel/make: 3.82-r5::gnome sys-kernel/linux-headers: 3.6 (virtual/os-headers) sys-libs/glibc: 2.16.0 Repositories: gentoo cross gnome x11 gentoo_fixes newstuff ACCEPT_KEYWORDS="* amd64 ~*" ACCEPT_LICENSE="* -@EULA" CBUILD="x86_64-pc-linux-gnu" CFLAGS="-march=barcelona -O2 -pipe -flto" CHOST="x86_64-pc-linux-gnu" CONFIG_PROTECT="/etc /usr/share/gnupg/qualified.txt" CONFIG_PROTECT_MASK="${EPREFIX}/etc/gconf /etc/ca-certificates.conf /etc/dconf /etc/env.d /etc/fonts/fonts.conf /etc/gconf /etc/gentoo-release /etc/revdep-rebuild /etc/sandbox.d /etc/terminfo" CXXFLAGS="-march=barcelona -O2 -pipe -flto" DISTDIR="/mnt/distfiles" FCFLAGS="-O2 -pipe" FEATURES="distlocks fixlafiles parallel-fetch protect-owned sandbox sfperms strict unmerge-orphans" FFLAGS="-O2 -pipe" GENTOO_MIRRORS="http://distfiles.gentoo.org" LDFLAGS="-Wl,-O1 -Wl,--as-needed -Wl,--hash-style=gnu" LINGUAS="en" MAKEOPTS="" PKGDIR="/var/tmp/packages" PORTAGE_CONFIGROOT="/" PORTAGE_RSYNC_OPTS="--recursive --links --safe-links --perms --times --compress --force --whole-file --delete --stats --human-readable --timeout=180 --exclude=/distfiles --exclude=/local --exclude=/packages" PORTAGE_TMPDIR="/var/tmp" PORTDIR="/mnt/portage" PORTDIR_OVERLAY="/mnt/portage/local/cross /mnt/portage/local/gnome /mnt/portage/local/x11 /mnt/portage/local/gentoo_fixes /mnt/portage/local/newstuff" SYNC="rsync://rsync.namerica.gentoo.org/gentoo-portage" USE="3dnow 3dnowext X a52 aac alsa amd64 avahi bluetooth boost bzip2 cairo consolekit cracklib css cups curl custom-cflags custom-cpuopts custom-optimization cxx dbus dts dvd eds egl encode exceptions expat fbcon ffmpeg flac fortran gif glib gnome gnome-keyring gnome-online-accounts gpm gstreamer gtk gtk3 icu introspection ipv6 jit jpeg lcms libnotify lzma mmx mmxext mp3 nautilus ncurses networkmanager nls nptl offensive ogg opengl openmp pam pango pcre png policykit postscript python3 qt3support qt4 readline samba speex spell sqlite sse sse2 sse3 ssl svg tcl theora threads tiff udev unicode usb vala vorbis x264 xcb xorg xv xvid zeroconf zlib" CURL_SSL="openssl" ELIBC="glibc" INPUT_DEVICES="evdev synaptics" KERNEL="linux" LINGUAS="en" PYTHON_TARGETS="python3_2" RUBY_TARGETS="ruby19" USERLAND="GNU" VIDEO_CARDS="intel nouveau radeon" Unset: CPPFLAGS, CTARGET, EMERGE_DEFAULT_OPTS, INSTALL_MASK, LANG, LC_ALL, PORTAGE_BUNZIP2_COMMAND, PORTAGE_COMPRESS, PORTAGE_COMPRESS_FLAGS, PORTAGE_RSYNC_EXTRA_OPTS, USE_PYTHON
Created attachment 329896 [details] build.log
It is not a bug in dev-python/notify-python. The problem looks similarly to bug #196243. Also disable -flto and rebuild libtool (bug #324107).
Sorry about the noise. Last time I tried removing -flto from notify-python, it didn't make any difference, but this time it worked. When I read through bug 196243. I noticed comment 20 mentions the exact problem I was having : global_symbol_pipe="" I just don't understand why -flto affects how the package's libtool is generated.
(In reply to comment #3) > I just don't understand why -flto affects how the package's libtool is > generated. An educated guess would be that one or more configure test go havoc with -flto. For gtkglext you'll run into a similar problem trying to compile with -flto. The problem is that most symbols of test programs are optimized out. I.e. the tests are too simple. A symbol not marked __attribute__((used)) is unlikely to make it into a generated object or binary file during ac_compile or ac_link if gcc -flto determines that "it is not used". Most of the tests generated by configure do not produce runnable executables, but gcc -flto optimizes for this szenario. The question is: is the symbol needed for execution of the compilation unit or binary? is there a dependency on the symbol? If not, it's optimized/thrown out of the generated file. The compiler does not know that configure is just probing it with some little test code, some object files, etc. - you'll have to tell it. The following patch makes the configure generated libtool of gtkglext work, if -flto is used to compile the test programs. Something similar will work for this bug as well. I've not tried, but it's possible that the patch will apply to this pkg's configure without modification (but with offset fuzziness). --- gtkglext-1.2.0/configure.orig 2013-05-07 03:48:34.987285143 +0200 +++ gtkglext-1.2.0/configure 2013-05-07 03:50:53.803539987 +0200 @@ -5946,11 +5946,9 @@ lt_cv_sys_global_symbol_to_c_name_address="sed -n -e 's/^: \([^ ]*\) $/ {\\\"\1\\\", (lt_ptr) 0},/p' -e 's/^$symcode* \([^ ]*\) \([^ ]*\)$/ {\"\2\", (lt_ptr) \&\2},/p'" ;; linux*) - if test "$host_cpu" = ia64; then symcode='[ABCDGIRSTW]' lt_cv_sys_global_symbol_to_cdecl="sed -n -e 's/^T .* \(.*\)$/extern int \1();/p' -e 's/^$symcode* .* \(.*\)$/extern char \1;/p'" lt_cv_sys_global_symbol_to_c_name_address="sed -n -e 's/^: \([^ ]*\) $/ {\\\"\1\\\", (lt_ptr) 0},/p' -e 's/^$symcode* \([^ ]*\) \([^ ]*\)$/ {\"\2\", (lt_ptr) \&\2},/p'" - fi ;; irix* | nonstopux*) symcode='[BCDEGRST]' @@ -6006,8 +6004,8 @@ #ifdef __cplusplus extern "C" { #endif -char nm_test_var; -void nm_test_func(){} +char nm_test_var __attribute__((used)); +void __attribute__((used)) nm_test_func(){} #ifdef __cplusplus } #endif
Created attachment 347560 [details, diff] configure-libtool-lto.patch untested patch working for a similar problem in another build
(In reply to comment #5) > untested patch working for a similar problem in another build If there are lot's of lto-unsave test programs in a given configure, an alternative is to sed -i configure to weave in "__attribute__((used))" where necessary. See https://bugs.gentoo.org/attachment.cgi?id=345856&action=diff for a working example.
Created attachment 347682 [details, diff] notify-python-0.1.1-r3.ebuild.patch Non-invasive patch to enable lto compiles. This is tested to work and is identical to the fix for gtkglext, see bug #468958 Please change the title of this bug fromr r2 to r3, thanks.
*** This bug has been marked as a duplicate of bug 463348 ***
