qtscriptgenerator @0.2.0_0: build error (command execution failed)

[victor.tran1@…]

I'm trying to install Amarok, but qtscriptgenerator fails to build properly.

OS X Yosemite 10.10.2

[victor.tran1@…]

Log file

[victor.tran1@…]

Cc Me!

[larryv (Lawrence Velázquez)]

Thanks for the report. In future tickets, please Cc any port maintainers (port info --maintainers qtscriptgenerator)

[RJVB (René Bertin)]

Fails for me too on OS X 10.9 . I'm suspecting a syntax incompatibility with clang++ . Looking into it.

[RJVB (René Bertin)]

Got a bit further, but still blocked. Comparing to a build on Linux, something must go wrong generating the code for QtScriptFutureIterator. On linux this class gets a well-filled implementation, on OS X the corresponding .cpp file remains in skeletal form. I haven't tried to understand why. The attached patch basically removes the QtScriptFutureIterator class to avoid the whole issue, but then the build fails later on because of missing targets (which I doubt are related to my patch):

make[2]: Entering directory `.../work/qtscriptgenerator-src-0.2.0/qtbindings/qtscript_gui'
make[2]: *** No rule to make target `../../generated_cpp/com_trolltech_qt_gui/plugin.cpp', needed by `plugin.o'.  Stop.
make[2]: Leaving directory `.../work/qtscriptgenerator-src-0.2.0/qtbindings/qtscript_gui'
make[1]: *** [sub-qtscript_gui-all] Error 2
make[1]: *** Waiting for unfinished jobs....
make[2]: Entering directory `.../work/qtscriptgenerator-src-0.2.0/qtbindings/qtscript_network'
make[2]: *** No rule to make target `../../generated_cpp/com_trolltech_qt_network/plugin.cpp', needed by `plugin.o'.  Stop.
make[2]: Leaving directory `.../work/qtscriptgenerator-src-0.2.0/qtbindings/qtscript_network'
make[2]: Entering directory `.../work/qtscriptgenerator-src-0.2.0/qtbindings/qtscript_core'
make[2]: Nothing to be done for `all'.
make[2]: Leaving directory `.../work/qtscriptgenerator-src-0.2.0/qtbindings/qtscript_core'
make[1]: *** [sub-qtscript_network-all] Error 2
make[2]: Entering directory `.../work/qtscriptgenerator-src-0.2.0/qtbindings/qtscript_opengl'
make[2]: *** No rule to make target `../../generated_cpp/com_trolltech_qt_opengl/plugin.cpp', needed by `plugin.o'.  Stop.

Hopefully Michaelld can get this a working. In the meantime, Amarok really isn't the best mediaplayer around O:-)

[michaelld (Michael Dickens)]

I see those errors too; haven't had time to look into it. QtScript Generator is pretty old now, and it's really not clear where it's repo is being stored. It's still necessary, though, so we'll figure something out ...

[RJVB (René Bertin)]

I see that the port is distributable, so what do the buildbots say (as in, on what OS X versions does the port build, and with which compiler)?

[macports@…]

Cc Me!

[macports@…]

Just a note that the ​Amarok homepage OS X download is just instructions to… install this port. They don’t have a built package available there.

[RJVB (René Bertin)]

Also not that that page is only a few months short of 3 years old and was written in a time gcc was still the default compiler. OTOH, someone (jrm) must have gotten amarok to build a bit less than a year ago.

I'd all forgotten about this ticket, and through a funny quirk of fate had an itch yesterday to look into the requirements for building Clementine through/against MacPorts. It's based on Amarok, but without all the KDE dependencies (and probably a better player).

[michaelld (Michael Dickens)]

If you can figure out how to coerce QtSG to build again RJVB, you'll make a lot of folks happy! I think it'll take longer than checking over all qt4-mac dependent ports to make sure the libexec install prefix works; but, maybe I'm wrong ;)

[RJVB (René Bertin)]

Evidently, checking over all qt4 dependent ports could have taken time for the build bots only and I'm pretty sure you'd have found less breakage if you'd simply taken my qt4-mac-devel port rather than deciding to dump all of Qt under libexec.

Getting qtscriptgenerator to build again isn't exactly high on my priority list as I don't run anything that depends on it myself (and memories of the last times I made altruist time investments in MacPorts are a bit too fresh). I could do another comparison with Linux, making sure this time I use clang there too, we'll see if I think of that the next time I don't know what else to do.

[michaelld (Michael Dickens)]

RJVB: In a prior comment you mention a patch that got you somewhere. Would you attach it to this ticket for us to try?

[michaelld (Michael Dickens)]

Has duplicate #45907. I'm moving that one here, since this ticket is cleaner.

[RJVB (René Bertin)]

It's a patch I made myself. I thought I'd attached it already (apparently got distracted before I could do so). It just removes the whole QtScriptFutureIterator class. I don't have access to the latest version right now, but I suppose it hasn't changed (much) w.r.t. the one in git: ​https://github.com/RJVB/macstrop/blob/master/devel/qtscriptgenerator/files/remove-QtScriptFutureIterator.diff

[michaelld (Michael Dickens)]

Ah; OK; NP. Thanks.

[RJVB (René Bertin)]

portfile diff

[RJVB (René Bertin)]

The above 4 diffs allow the port to build for me, against my own concurrent Qt4 port (which installs only partly into libexec/qt4). No patches to phonon are required in my approach.

The remove-QtScriptFutureIterator.diff patch fixes an issue I ran into a while back; I haven't checked whether it is still required with the current fixes to the headerfile search logic.

Similarly, I haven't yet checked whether the patch-include-everything.diff is indeed required (it's a known fix that lets qtsg find KDE's phonon).

[persramd@…]

Cc Me!

[michaelld (Michael Dickens)]

Maybe I'm doing something wrong, but your patches still fail for me on 10.11 / Xcode 7 (latest release or beta of both). I'm attaching my full diff & log file for reference.

[michaelld (Michael Dickens)]

BTW> Please don't hardwire /opt/local into patches. You know what to do .... ;)

[RJVB (René Bertin)]

The annoying thing with qmake is that it's such a hassle to get it to use a different compiler on OS X; I haven't managed to import macports-clang-3.7 . However, when I compile just the failing file with that compiler it works fine.

You don't happen to have a VM with older OS X/Xcode versions around?

[mkae (Marko Käning)]

See also duplicate #52346!

[iMiKED (iMiKE)]

Cc Me!

[michaelld (Michael Dickens)]

Unless a miracle occurs, this port is going nowhere. The programming is obtuse / unclear. Even Rene's mods don't work in general (mostly just for his specific case). I have no spare time right now to look into this issue. I welcome patches on the main SVN trunk / GIT master.

[RJVB (René Bertin)]

Michael, could you at least find the time to tell me what exactly in my mods doesn't work for the generic case? I have a hunch it has to do with my mods that points INCLUDEPATH to ${prefix}/include/qt4 instead of to ${qt_includes_dir}?

Also, where are the phonon headers installed in the generic case, hopefully nice and simply under ${prefix}/include/phonon ?

Edit: for anyone interested, the portdir with my mods is here: ​https://github.com/RJVB/macstrop/tree/master/devel/qtscriptgenerator

The easiest way to test it is to archive your original port dir qtscriptgenerator, and replace it with the contents from my repo.

[michaelld (Michael Dickens)]

Did I never get back to you? Hmm ... well, whatever & my bad if I didn't. Are the patches the same for 10.12 as they are attached here? I'll give it a go & report back. Will your QTSG work with the MP trunk directly?

[RJVB (René Bertin)]

Sorry, I had an extended Sunday lunch :)

I have to presume that 10.12 doesn't change anything, or should I say that this old qtscripgenerator code doesn't do anything that needs to be adapted for 10.12 . I can't test that at the moment. And it certainly is my intention that this port can be built to work with your port:qt4-mac. Today's change should have removed the last hardcoded assumption about where things are installed, but I may be overlooking something. It's so long ago that I figured out how to get this to build that I have mostly forgotten the details :-/

[michaelld (Michael Dickens)]

Works for me. Committed in r153140. I did remove some of the patches, since they were unused at least for MacPorts purposes. Thanks!

[RJVB (René Bertin)]

Glad to see at least some of my efforts get to serve the greater good :)

[iMiKED (iMiKE)]

Excuse me, but how to get & compile this fixed updated version? Macports selfupdate does nothing - still can't compile.