Error log is attached.
Created attachment 5780 [details] error log of emerge
Additional info: 0.49 is out. 0.48 can no longer be emerged, because the download is away. However, the compile problem is not fixed.
Created attachment 6053 [details, diff] potential fix for the kwin Bluecurve build The problem is an old .moc file--making a new .moc file fixes the problem.
Another strange thing: if I enter the art/qt/Bluecurve dir from the command line and issue 'make' it builds just fine. It's been too long since I've crawled through a configure-generated makefile, I swear. :-D Also, I noticed that using the patch I submitted may not be necessary--it may just be necessary to remove the .moc files altogether. There's a moc build target in Makefile.am in art/qt/Bluecurve, so perhaps it's possible. Maybe a case of sloppy package building on Redhat's part? Hm. Then again, I can't really talk ill of them, bc I'm not really much of a coder at all. ;-D
OK, here's what's weird. If I build from the toplevel, here's what happens right before the build bombs out with an error: /bin/sh ../../../libtool --mode=compile c++ -DHAVE_CONFIG_H -I. -I. -I../../.. -I/usr/kde/3/lib/qt3/include -DQT_THREAD_SUPPORT -D_REENTRANT -Wnon-virtual-dtor -Wno-long-long -Wbad-function-cast -Wundef -Wall -pedantic -W -Wpointer-arith -Wmissing-prototypes -Wwrite-strings -ansi -D_XOPEN_SOURCE=500 -D_BSD_SOURCE -Wcast-align -Wconversion -O2 -g -O2 -fno-exceptions -fno-check-new -c plugin.cpp c++ -DHAVE_CONFIG_H -I. -I. -I../../.. -I/usr/kde/3/lib/qt3/include -DQT_THREAD_SUPPORT -D_REENTRANT -Wnon-virtual-dtor -Wno-long-long -Wbad-function-cast -Wundef -Wall -pedantic -W -Wpointer-arith -Wmissing-prototypes -Wwrite-strings -ansi -D_XOPEN_SOURCE=500 -D_BSD_SOURCE -Wcast-align -Wconversion -O2 -g -O2 -fno-exceptions -fno-check-new -Wp,-MD,.deps/plugin.pp -c plugin.cpp -fPIC -DPIC -o plugin.lo plugin.cpp:6:26: qstyleplugin.h: No such file or directory Now, here's what happens if I enter that dir and fire off a 'make' (the build, BTW, does just fine:) /bin/sh ../../../libtool --mode=compile c++ -DHAVE_CONFIG_H -I. -I. -I../../.. -I/usr/qt/3/include -DQT_THREAD_SUPPORT -D_REENTRANT -Wnon-virtual-dtor -Wno-long-long -Wbad-fu nction-cast -Wundef -Wall -pedantic -W -Wpointer-arith -Wmissing-prototypes -Wwrite-strings -ansi -D_XOPEN_SOURCE=500 -D_BSD_SOURCE -Wcast-align -Wconversion -O2 -g -O2 -fno-exc eptions -fno-check-new -c plugin.cpp c++ -DHAVE_CONFIG_H -I. -I. -I../../.. -I/usr/qt/3/include -DQT_THREAD_SUPPORT -D_REENTRANT -Wnon-virtual-dtor -Wno-long-long -Wbad-function-cast -Wundef -Wall -pedantic -W -Wpo inter-arith -Wmissing-prototypes -Wwrite-strings -ansi -D_XOPEN_SOURCE=500 -D_BSD_SOURCE -Wcast-align -Wconversion -O2 -g -O2 -fno-exceptions -fno-check-new -Wp,-MD,.deps/plugin .pp -c plugin.cpp -fPIC -DPIC -o plugin.lo Now, what seems odd to me is that there would be a difference--as I say, I'm not really a coder and GNU's build utils are about as opaque to me as they could be. Building from toplevel wants to include dirs *that are wrong.* That's the only problem. I suppose I'm too damn tired to see the solution, but perhaps the -I/usr/kde/3/lib/qt3/include is a RHism? Dang that RH, but again I really can't blame 'em because it is a set of themes designed for their distro. ;-D I shall look at it again after I get a few hours sleep.
One more thing before I go to bed, and again, I must just be too darn sleepy to see the obvious answer. For whatever reason, the generated Makefile in redhat-artwork-0.48/art/qt contains the following weirdness: prefix = /usr/kde/3 exec_prefix = ${prefix} *snip* libdir = ${exec_prefix}/lib *snip* QTDIR = $(libdir)/qt3 it would seem that QTDIR is pointing at /usr/kde/3/qt3. In all_includes, though, the include dir is right. If you look at the Makefile in redhat-artwork-0.48/art/qt, INCLUDES is set to INCLUDES = -I$(QTDIR)/include. Weird stuff. The reason it builds, then, when one goes to the dir, is because QTDIR is set correctly as an environment variable, and the Makefile in the previous dir incorrectly overrides this with an incorrect setting. I'm going to fix the Makefile.am, submit a patch, and go to bed. ;-D .
Created attachment 6054 [details, diff] fix for braindead automake file Removes QTDIR from Makefile.am in art/qt I suppose I should have checked 0.49 to see if it still suffers from this brainfart as well. :-D
Created attachment 6055 [details] A real fix :-} Sorry about this; last one, I promise. :-D Basically what was happening was that there are old .moc files lying around that need to be replaced. This edited ebuild replaces those .moc files. I've already tested the build--it works. The only problem now is that there are still sandbox violations.
Erm, hello? Anyone home? :-D I appreciate the fact that you're a for-free developer (for-free developers) but really, hey, if you're not maintaining the package anymore (not that I'd want to do it) would you consider turning it over to someone else? Not meaning to be a smartass; Actually, I was just bragging a while back about the joys of Gentoo, and used redhat-artwork as an example of what was so great about Gentoo. The fact that software makes it into Gentoo so darn quick, in other words. Please don't make a liar out of me. ;-D If the last patch I submitted sucked, please tell me so. :D
Sorry, I'm inactive and these bugs should have been reassigned to someone else. Bouncing to bug-wranglers.
*** Bug 11773 has been marked as a duplicate of this bug. ***
Created attachment 6599 [details, diff] Yet Another Patch (Possibly Redundant) fixes the same problem the automake patch and the moc patch fixes.. just an alternative way of doing it (well i hope it works anyway)
Shane, Ben: does the last patch work? comment #12? or which patch does work?
mmm sorry about the confusion i think i just worked out whats going on i think we need both mine and shanes fixes mine fixes a different bug (i should of read the error log before posting :( ) i cant seem to compile with just shane's fixes but my fix is good enough for me to compile most people will probably be able to compile with shane's fixes but people using kde-cvs ebuilds probably wont be able to compile with just shanes fixes (or people compiling redhat-artwork in the future will have a problem) just to clarify.. mine fixes some hard coded paths in some of the Makefile.am ie redhat-artwork-0.49/art/qt/Makefile.am QTDIR = $(libdir)/qt-3.1 and shane's patch regenerates the moc files which was the problem in this bug report also you may want to bump the RH_EXTRAVERSION="2" to 3 in shane's ebuild to make it identical to whats in portage at the moment
*** Bug 11712 has been marked as a duplicate of this bug. ***
*** Bug 11831 has been marked as a duplicate of this bug. ***
i have just commitet version 0.53 as testing. it doesn't compile the kde-part for me, tried several of the fixes posted here. however, the moc-thing seems to be fixed on the redhat-side. will have to look at it again later.
Hanno, what error are you getting on 0.53? I'm getting a weird list of error messages in the Kwin dir (mostly just warnings): g++ -DHAVE_CONFIG_H -I. -I. -I../../../.. -I/usr/kde/3/include -I/usr/qt/3/include -I/usr/X11R6/include -I../../../.. -DQT_THREAD_SUPPORT -D_REENTRANT -Wnon-virtual-dtor -Wno-long-long -Wbad-function-cast -Wundef -Wall -pedantic -W -Wpointer-arith -Wmissing-prototypes -Wwrite-strings -ansi -D_XOPEN_SOURCE=500 -D_BSD_SOURCE -Wcast-align -Wconversion -O2 -O2 -pipe -fno-exceptions -fno-check-new -c redhatclient.cpp -fPIC -DPIC -o redhatclient.lo redhatclient.cpp: In member function `void KWinInternal::RedHatHandler::createPixmaps()': redhatclient.cpp:279: warning: assignment to `int' from `double' redhatclient.cpp:279: warning: argument to `int' from `double' redhatclient.cpp: In member function `virtual void KWinInternal::RedHatClient::paintEvent(QPaintEvent*)': redhatclient.cpp:1124: warning: assignment to `int' from `double' redhatclient.cpp:1124: warning: argument to `int' from `double' redhatclient.cpp:977: warning: unused variable `int offset' redhatclient.cpp:995: warning: unused variable `int leftFrameStart' redhatclient.cpp:999: warning: unused variable `int rightOffset' redhatclient.cpp:1001: warning: unused variable `QColor c2' redhatclient.cpp: In member function `void KWinInternal::RedHatClient::doShape()': redhatclient.cpp:1260: warning: unused variable `QRect r' redhatclient.cpp:1263: warning: unused variable `int x2' redhatclient.cpp:1264: warning: unused variable `int y2' redhatclient.cpp: In member function `void KWinInternal::RedHatClient::menuButtonPressed()': redhatclient.cpp:1415: no matching function for call to ` KWinInternal::Workspace::showWindowMenu(QPoint, KWinInternal::RedHatClient* const)' make[5]: *** [redhatclient.lo] Error 1 make[5]: Leaving directory `/var/tmp/portage/redhat-artwork-0.53/work/redhat-artwork-0.53/art/kde/kwin/Bluecurve' make[4]: *** [all-recursive] Error 1 make[4]: Leaving directory `/var/tmp/portage/redhat-artwork-0.53/work/redhat-artwork-0.53/art/kde/kwin' make[3]: *** [all-recursive] Error 1 make[3]: Leaving directory `/var/tmp/portage/redhat-artwork-0.53/work/redhat-artwork-0.53/art/kde' make[2]: *** [all-recursive] Error 1 make[2]: Leaving directory `/var/tmp/portage/redhat-artwork-0.53/work/redhat-artwork-0.53/art' make[1]: *** [all-recursive] Error 1 make[1]: Leaving directory `/var/tmp/portage/redhat-artwork-0.53/work/redhat-artwork-0.53' make: *** [all-recursive-am] Error 2 I'm at a loss; I'll do a little bumbling around to see what might be up later.
Created attachment 6904 [details, diff] Old Function fix Whats happening is that the program is using the new showWindowMenu function in kwin.h. By using this function they fixed some weird errors that were occuring with the latest kde revisions because clientPopup(Client *cl) is becoming depreciated. (I'm getting segmentation faults in some situations with redhat-artwork-0.49) Patch changes configure.in & acconfig.h adds a test to see if the new showWindowMenu function is defined in kwin.h and if it is defines HAVE_SHOWWINDOW in config.h redhatclient.cpp checks if HAVE_SHOWWINDOW is defined before using the new function, if it isn't then use the old one you will need to run autoconf and autoheader before running configure Problems: I have a cvs kde installed and if i use the old function it segmentation faults in some cases. I've looked at the difference between a 3.0.0 kwin.cpp and cvs kwin.cpp and I believe it should be fine. This behaviour also occured in redhat-artwork-0.49.
*** Bug 13336 has been marked as a duplicate of this bug. ***
I just added redhat-artwork-0.58 as ~x86 (unstable) and it compiles the kde-part for me. So it seems these issues have been fixed upstream. Can please everyone who noticed compile errors with redhat-artwork test 0.58 if it a) compiles with USE-var kde enabled b) works and report the results here. If yes, I would like to have this marked stable soon.
I still cannot emerge this. I'll try disabling sandbox and see if it works. ve.so /var/tmp/portage/redhat-artwork-0.58/image/usr/kde/3.1/lib/kwin_bluecurve.so >>> Completed installing into /var/tmp/portage/redhat-artwork-0.58/image/ --------------------------- ACCESS VIOLATION SUMMARY ---------------------------LOG FILE = "/tmp/sandbox-redhat-artwork-0.58-26198.log" open_wr: /usr/qt/3/etc/settings/.qtrc.lock open_wr: /usr/qt/3/etc/settings/.qtrc.lock open_wr: /usr/qt/3/etc/settings/.qtrc.lock open_wr: /usr/qt/3/etc/settings/.qtrc.lock
Disabling sandbox (editing /etc/make.globals and removing "sandbox" from FEATURES) causes this compile to work for me, fail as above otherwise -Steve
It is a known fact that uic and moc, which are used when building qt/kde apps, like to write in ~/.qt and ~./kde. The redhat-artwork ebuild should include this fix from kde.eclass: # fix the sandbox errors "can't writ to .kde or .qt" problems. # this is a fake homedir that is writeable under the sandbox, so that the build process # can do anything it wants with it. REALHOME="$HOME" mkdir -p $T/fakehome/.kde mkdir -p $T/fakehome/.qt export HOME="$T/fakehome" # things that should access the real homedir [ -d "$REALHOME/.ccache" ] && ln -sf "$REALHOME/.ccache" "$HOME/" I don't want to get involved with it though, beacuse ther' all the non-kde stuff in there... Also the latset ebuild can't fetch its sources again, because version 0.62-1 has been released by redhat. Hanno, could you go on handling this please?
I have now commited a 0.63-version to cvs. But it still doesn't work for the kde-part. It seems that it compiles fine, but tries to install into the wrong directories. Fixes are welcome!
I'm committing an 0.63-r1 which should fix that. Please test.
Are there still any compile problems with 0.65? If not, I would like to mark it stable and close this bug.
I'm marking this fixed for now, if anyone still has compile-problems, please reopen this bug.
