Current Git broken?

• Apr 15, 2016 - 12:37

1. hope this is the right forum for my problem.

2.
I'm compiling mscore-unstable, using the git source tree. Until yesterday, the make process was performed flawlessly. But since todya, mscore won't compile:

gmake stops after I made "git pull":

--------------------------------------------------------------------------------------

git pull

...
...

Generating moc_textproperties.cpp
Generating moc_textstyledialog.cpp
Generating moc_texttools.cpp
Generating moc_timedialog.cpp
Generating moc_timesigproperties.cpp
Generating moc_transposedialog.cpp
Generating moc_tremolobarcanvas.cpp
Generating moc_tremolobarprop.cpp
Generating moc_tupletdialog.cpp
Generating moc_updatechecker.cpp
Generating moc_uploadscoredialog.cpp
Generating moc_voltaproperties.cpp
Generating moc_waveview.cpp
Generating moc_workspace.cpp
make[3]: Leaving directory '/muse/MuseScore/build.release'
[ 0%] Built target mscore_automoc
make[2]: Leaving directory '/muse/MuseScore/build.release'
Makefile:149: recipe for target 'all' failed
make[1]: *** [all] Error 2
make[1]: Leaving directory '/muse/MuseScore/build.release'
Makefile:46: recipe for target 'release' failed
gmake: *** [release] Error 2

Anybody has a similar problem? My platform: Fedora 23, x86_64.

All these steps were performed flawlessly until yesterday.

Comments

• Apr 15, 2016 - 13:10

same problem in Fedora 24! Installed qt5-packages:

rpm -qa qt5*
qt5-qtscript-devel-5.6.0-3.fc24.x86_64
qt5-qtlocation-5.6.0-3.fc24.x86_64
qt5-qtquick1-5.6.0-3.fc24.x86_64
qt5-qttools-libs-help-5.6.0-3.fc24.x86_64
qt5-qttools-libs-designercomponents-5.6.0-3.fc24.x86_64
qt5-qtbase-gui-5.6.0-9.fc24.x86_64
qt5-qtwebchannel-5.6.0-3.fc24.x86_64
qt5-qtxmlpatterns-5.6.0-4.fc24.x86_64
qt5-designer-5.6.0-3.fc24.x86_64
qt5-qtdeclarative-devel-5.6.0-4.fc24.x86_64
qt5-qtwebkit-5.6.0-4.fc24.x86_64
qt5-qttools-libs-designer-5.6.0-3.fc24.x86_64
qt5-qtquickcontrols-5.6.0-3.fc24.x86_64
qt5-qtbase-common-5.6.0-9.fc24.noarch
qt5-qtscript-5.6.0-3.fc24.x86_64
qt5-rpm-macros-5.6.0-9.fc24.noarch
qt5-qtsvg-devel-5.6.0-3.fc24.x86_64
qt5-qtwebkit-devel-5.6.0-4.fc24.x86_64
qt5-qttools-libs-clucene-5.6.0-3.fc24.x86_64
qt5-qtbase-5.6.0-9.fc24.x86_64
qt5-qttools-5.6.0-3.fc24.x86_64
qt5-qtbase-devel-5.6.0-9.fc24.x86_64
qt5-qhelpgenerator-5.6.0-3.fc24.x86_64
qt5-qttools-static-5.6.0-3.fc24.x86_64
qt5-qtdeclarative-5.6.0-4.fc24.x86_64
qt5-qtsvg-5.6.0-3.fc24.x86_64
qt5-linguist-5.6.0-3.fc24.x86_64
qt5-qtxmlpatterns-devel-5.6.0-4.fc24.x86_64
qt5-qtsensors-5.6.0-3.fc24.x86_64
qt5-qdoc-5.6.0-3.fc24.x86_64
qt5-qtquick1-devel-5.6.0-3.fc24.x86_64
qt5-qttools-common-5.6.0-3.fc24.noarch
qt5-qttools-devel-5.6.0-3.fc24.x86_64

Reply

• Apr 15, 2016 - 15:51

That single huge PR leaves MuseScore uncompilable even on Debian Testing. That's, errr, unfortunate. Qt5.6 is only available in Debian Experimental (maked rc-buggy).
Did this have to happen in one big chunk? (That's often a sign of unelegant version control handling ).

Reply

• Apr 15, 2016 - 18:04

Yes, but this single PR/commit combines a lot of different features/changes. This is exactly how git is not supposed to be used. In case of a bug (an there will be some) git bisect will just shamefully point at that huge hunk of changes ... not very helpful.

BTW - I know how to build against current Qt5, and in theory it does work, except:

{syntaxhighlighter shell}

[ 0%] Generating all.h.gch
In file included from /opt/Qt/Qt5.6.0/5.6/gcc_64/include/QtCore/qnamespace.h:37:0,
from /opt/Qt/Qt5.6.0/5.6/gcc_64/include/QtCore/qobjectdefs.h:41,
from /opt/Qt/Qt5.6.0/5.6/gcc_64/include/QtCore/qobject.h:40,
from /opt/Qt/Qt5.6.0/5.6/gcc_64/include/QtCore/qabstractanimation.h:37,
from /opt/Qt/Qt5.6.0/5.6/gcc_64/include/QtCore/QtCore:4,
from /opt/Qt/Qt5.6.0/5.6/gcc_64/include/QtGui/QtGuiDepends:3,
from /opt/Qt/Qt5.6.0/5.6/gcc_64/include/QtGui/QtGui:3,
from all.h:33:
/opt/Qt/Qt5.6.0/5.6/gcc_64/include/QtCore/qglobal.h:1090:4: error: #error "You must build your code with position independent code if Qt was built with -reduce-relocations. " "Compile your code with -fPIC (-fPIE is not enough)."
# error "You must build your code with position independent code if Qt was built with -reduce-relocations. "\
^
CMakeFiles/mops2.dir/build.make:60: recipe for target 'all.h.gch' failed
make[2]: *** [all.h.gch] Error 1
CMakeFiles/Makefile2:200: recipe for target 'CMakeFiles/mops2.dir/all' failed
make[1]: *** [CMakeFiles/mops2.dir/all] Error 2
Makefile:149: recipe for target 'all' failed
make: *** [all] Error 2

{/syntaxhighlighter}

Is this really up for the master branch?

Reply

• Apr 15, 2016 - 18:15

Thank you for the git lesson...

The master branch currently build and run successfully on Mac OSX, Windows and Ubuntu 12.04 Precise (our CI environment). I can't imagine it can't build on Debian Testing. Where is your Qt from?

Reply

• Apr 15, 2016 - 18:24

Fresh download from the official Qt website (the offline installer qt-opensource-linux-x64-5.6.0.run).

The reported problem seems to be a CMake problem (I'm using the CMake Support as provided by the Qt installation). The '-fPIC' flag is missing in CMAKE_CXX_FLAGS and CMAKE_CXX_FLAGS_DEBUG).

Reply

• Apr 15, 2016 - 18:38

If you're running Windows you might find my 5.4 to 5.6 upgrade docs/images helpful. See here: https://musescore.org/en/node/106326

I just realized it left out the steps I took with my env variables, like PATH anc CXX. Now I have to add those... but they are the same steps as in the standard compile instructions.

I have just successfully built and deployed today's MuseScore/master as of a few minutes ago.

Reply

• Apr 15, 2016 - 20:18

I cannot reproduce the result of your attached image.
I receive presently an expected result (Tab for lute 13-course). See attachment.
Feel free to provide precise steps in the Issue Tracker for reproduce what you get (the "broken tab")

Attachment Size
lute1.jpg 24.08 KB

Reply

• Apr 15, 2016 - 20:52

Newly created tabs seem to work - opening some old samples created that broken rendering. I'm trying to find that sample but currently most of my samples trigger a segfault in libmscore/staff.cpp:748:

{syntaxhighlighter shell}
Program received signal SIGSEGV, Segmentation fault.
Ms::Staff::lines (this=this@entry=0x0) at /usr/local/src/MuseScore/libmscore/staff.cpp:748
748 return _staffType.lines();
(gdb) l
743 // lines
744 //---------------------------------------------------------
745
746 int Staff::lines() const
747 {
748 return _staffType.lines();
749 }
750
751 //---------------------------------------------------------
752 // setLines

{/syntaxhighlighter}

Reply

• Apr 16, 2016 - 02:18

It might be a lot of different commits, but it's basically one big change - a radically changed layout algorithm that you'll be hearing much more about soon. So it's not necessarily a bad thing that any bug introduced by this points to that one commit.

Reply

• Apr 18, 2016 - 04:10

OS: Manjaro / Arch Linux
Problem: Compilation complained about Qt5WebEngineWidgets.
Everything worked fine until a few days ago.

I was able to fix this by installing qt5-webengine. For those who use pacman as their package manager:
sudo pacman -S qt5-webengine

Reply

Do you still have an unanswered question? Please log in first to post your question.