Cannot merge octave-3.8.2-r3 due to the following error: libtool: compile: i686-pc-linux-gnu-g++ -DHAVE_CONFIG_H -I. -I.. -I/usr/include/qt4 -I/usr/include/qt4/QtGui -I/usr/include/qt4 -I/usr/include/qt4/QtNetwork -I/usr/include/qt4 -I/usr/include/qt4/QtCore -I./qterminal/libqterminal -Isrc -I./src -I./src/m-editor -I./src/qtinfo -I../liboctave/cruft/misc -I../liboctave/array -I../liboctave/numeric -I../liboctave/numeric -I../liboctave/operators -I../liboctave/operators -I../liboctave/system -I../liboctave/util -I../libinterp -I../libinterp -I../libinterp/parse-tree -I../libinterp/parse-tree -I../libinterp/corefcn -I../libinterp/corefcn -I../libinterp/octave-value -I../libgnu -I../libgnu -O2 -march=pentium4 -pipe -mieee-fp -Wall -W -Wshadow -Wold-style-cast -Wformat -Wpointer-arith -Wwrite-strings -Wcast-align -Wcast-qual -O2 -march=pentium4 -pipe -pthread -c src/workspace-view.cc -fPIC -DPIC -o src/.libs/src_libgui_src_la-workspace-view.o ( echo '#ifdef HAVE_CONFIG_H'; echo '#include <config.h>'; echo '#endif'; moc -DHAVE_CONFIG_H -I. -I.. -O2 -march=pentium4 -pipe src/m-editor/file-editor-interface.h ) > src/m-editor/moc-file-editor-interface.cc-t ( echo '#ifdef HAVE_CONFIG_H'; echo '#include <config.h>'; echo '#endif'; moc -DHAVE_CONFIG_H -I. -I.. -O2 -march=pentium4 -pipe src/m-editor/file-editor-tab.h ) > src/m-editor/moc-file-editor-tab.cc-t moc: Invalid argument Usage: moc [options] <header-file> -o<file> write output to file rather than stdout -I<dir> add dir to the include path for header files -E preprocess only; do not generate meta object code -D<macro>[=<def>] define macro, with optional definition -U<macro> undefine macro -i do not generate an #include statement -p<path> path prefix for included file -f[<file>] force #include, optional file name -nn do not display notes -nw do not display warnings @<file> read additional options from file -v display version of moc Note: 'moc' is a symlink to 'qtchooser'. I have dev-qt/qtchooser-0_p20151008. I have Qt4, no Qt5.
Can I see log and CFLAGS? I had some trouble with octave and moc (basically with the -f switch of moc you can add a custom header, I had -fopenmp in CFLAGS, for some reason them were not filtered out and finished in the moc invocation. Result: #include "openmp" in some files) last week, I solved by emerging automoc. Maybe it should be added as dependency?
Oh I see that you have -fPIC, I bet this is driving it crazy.
Created attachment 471846 [details] Build log for sci-mathematics/octave-3.8.2-r3 I attach the build log for octave-3.8.2-r3. And here comes the strange thing: after posting the bug, I thought "well, actually I do have octave-3.8.2-r1 already and the reason I wanted to merge was some new version of ImageMagick that caused portage to lament about reverse dependencies. So let me re-compile exactly that version, which I *know* it works". I found the 3.8.2-r1 ebuild somewhere, put it into a local overlay and started. And now the strange thing: compilation stopped with exactly the same error at exactly the same position! This suggests that it has something to do with my Qt4, rather with octave, or the ebuild. But I don't remember touching Qt4, especially since I mask Qt5 and the package that installs moc has a version of 2015... To put it in perspective: some third party package changed (say, ImageMagick, with a new version, or new flag, I don't remember). Then portage complained about revdep-rebuild. I saw octave in the list and decided to re-merge. Now both a working (installed) ebuild (3.8.2-r1) and a slightly newer one (3.8.2-r3) stop at the same place with an error about moc, a 'Qt version chooser' with a version of 0_p20151008 - old enough and unchanged since 3.8.2-r1 was installed without problems last August... How can such a thing happen?
...ah, and -fPIC is not the problem, as you can see. I use the 'hardened toolchain', but not the hardened kernel (yet). Problems are very, very rare (mostly "asm operand has impossible constraints", which has to do with the 32bit architecture and can be resolved by a line use x86 && append-cflags $(test-flags-CC -fno-stack-check) in the ebuild, see bug #499996 for a discussion). The only change between last August and today is the kernel - from 3.16 to 4.9.25. This brought some minimal changes in the packages, namely elt-patches-20170317 sys-apps/portage-2.3.3 sys-devel/binutils-2.26.1 dev-util/perf-4.1.5-r1 app-shells/quoter-3.0_p2 app-shells/push-2.0 sys-kernel/genkernel-3.4.52.4-r2 app-portage/eix-0.32.4 and then sys-kernel/gentoo-sources-4.9.25 This was followed by: dev-perl/DBD-mysql-4.41.0 x11-libs/qscintilla-2.9.2-r1 dev-util/cmake-3.7.2 media-gfx/imagemagick-6.9.7.4 media-gfx/uniconvertor-2.0_pre379-r1 media-gfx/inkscape-0.91-r3 media-libs/svgalib-1.9.25-r4 app-emulation/virtualbox-modules-5.0.32 app-emulation/virtualbox-5.0.32 x11-drivers/nvidia-drivers-304.135 dev-php/pecl-imagick-3.4.1 and then octave with this problem. The only things that I could suspect in the above list are binutils, quoter, push and cmake. If I had to bet, I would bet on cmake - old version was cmake-3.5.2-r1. I would bet that a change in cmake between 3.5.2-r1 and 3.7.2 causes the working moc of qtchooser-0_p20151008 to break with the error in the log.
...and I do have dev-util/automoc-0.9.88-r1 installed, the current version. Just having it installed does not change anything, so it is probably unrelated. I still bet on cmake... :-)
Removed from tree, reopen if issue persists. commit 8fc2192f2c98e1de3f9667d4d968141c6df8d55c Author: David Seifert <soap@gentoo.org> Date: Sun Jun 11 23:02:50 2017 +0200 sci-mathematics/octave: Remove old
