[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#876131: qtbase-opensource-src FTCBFS: uses the build architecture toolchain



On 20/01/16 11:18, Dmitry Shachnev wrote:
> Hi Helmut!
> 
> On Thu, Jan 16, 2020 at 08:25:47PM +0100, Helmut Grohne wrote:
> > I can explain this. Please compare the relevant mkspecs:
> >
> > https://sources.debian.org/src/qtbase-opensource-src/5.12.5+dfsg-5/mkspecs/linux-g++/qmake.conf/
> > https://sources.debian.org/src/qtbase-opensource-src/5.12.5+dfsg-5/mkspecs/linux-aarch64-gnu-g++/qmake.conf/
> >
> > You'll quickly observe that they look much the same with one key
> > difference. The arm64 one has all the tools prefixed with the GNU
> > triplet. And that's precisely the property we need here. In principle,
> > could mechanically generate a similar mkspec for any other architecture.
> > It is much like a CMake toolchain file or a meson toolchain file. We
> > need it to tell QT which architecture to build for.
> >
> > Possibly we could patch a new linux-debian-g++ mkspec that sets up the
> > variables based on the dpkg-architecture environment variables. If we go
> > that route, make sure not to install that into a binary package. Prior
> > art: linux-oe-g++
> > (https://github.com/meta-qt5/meta-qt5/wiki/Building-with-OE).
> 
> I wonder if we can use the linux-g++ mkspec and pass QMAKE_CXX and similar
> variables to configure, just like we pass them to qmake for cross builds:
> 
> https://salsa.debian.org/qt-kde-team/qt/qtbase/blob/master/debian/qmake-cross-wrapper.in
> 
> configure just passes its options to qmake anyway:
> 
> https://sources.debian.org/src/qtbase-opensource-src/5.12.5+dfsg-5/configure/#L857

We should give this a try.

> Or maybe set external-hostbindir to /usr/lib/${DEB_BUILD_MULTIARCH}/qt5/bin,
> in which case qmake will be a symlink to qmake-cross-wrapper? Though thinking
> more about it, in this case the real qmake will get two -qtconf options, of
> which the latter (added by configure) will be wrong. So that will probably not
> work.

Indeed.


Reply to: