Granted, this is a silly combination of use flags, but the ebuild should still handle it gracefully rather than failing during configure. Fix is simple but I'll attach a patch anyway. Reproducible: Always Steps to Reproduce:
Created attachment 74510 [details, diff] Check for invalid use flag combination
How does it fail? need-kde() calls need-qt(), so the dependencies should be correct. In any case, I don't think an ebuild should die because of USE='kde -qt'. Rather, make it behave like USE='kde qt'.
(In reply to comment #2) > How does it fail? [[[ checking for qt3to4... NO checking if KDE3 is allowed... yes checking if QT3 is allowed... yes checking for qt3 includes... -I/usr/qt/3/include checking if threaded qt3 may be used... yes checking for qt3 libraries... not found checking for qt3 moc... /usr/qt/3/bin/moc checking for qt3 uic... /usr/qt/3/bin/uic configure: WARNING: QT3 is not explicitly disabled and I could not find some QT3 components (see which are missing in messages above). If you don't want to compile QT3 applications please use "--disable-qt3". checking if qt3 version is fully useable... no checking target system type... posix checking for xgettext... xgettext checking for locale.h... (cached) yes checking libintl.h usability... yes checking libintl.h presence... yes checking for libintl.h... yes checking for library containing gettext... none required checking for msgfmt... /usr/bin/msgfmt checking whether msgfmt understands --qt... yes checking whether this is an official release... no checking for compiler arguments... -march=athlon64 -O2 -pipe -Wall -Wdeclaration-after-statement checking docpath... /var/tmp/portage/homedir/apidoc checking if full docu should be created... no checking doxygen tag files... none checking whether local install mode is wanted... no configure: *** KDE3 is required for KDE-frontend "kbanking". Specify --with-frontends="cbanking g2banking qbanking" to build aqbanking without that frontend. configure: error: QT3 frontend is needed by KDE-frontend. ]]] > need-kde() calls need-qt(), so the dependencies should be correct. Well... > In any case, I don't think an ebuild should die because of USE='kde -qt'. > Rather, make it behave like USE='kde qt'. Absolutely not. Ebuilds have no business silently changing use flags. On the other hand, if you mean that it shouldn't have a qt use flag at all, then yes, that makes sense. :)
(In reply to comment #3) > > In any case, I don't think an ebuild should die because of USE='kde -qt'. > > Rather, make it behave like USE='kde qt'. > > Absolutely not. Ebuilds have no business silently changing use flags. On the > other hand, if you mean that it shouldn't have a qt use flag at all, then yes, > that makes sense. :) Yes they do. If the kde frontend needs the qt frontend, qt support is no longer optional and the use flag should thus be ignored.
So perhaps the smart thing to do is to build some logic into the kde-function eclass that fails emerge if you try to do a USE="-qt kde" combation. KDE herd, thoughts?
should be fixed now
