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

Bug#734677: approach



Dima Kogan <dima@secretsauce.net> writes:

> I'll make a patch for upstream when I get the chance. The patch I sent
> earlier let most headers remain in /usr/include, and moved just a single
> file to /usr/include/<triplet>. Is this reasonable?

Hi.

I just looked at the way the upstream QT sources ship pkgconfig and
qmake files. Both those sets of files are generated, it turns out. The
sources make an assumption that there's a single include directory. So
if we go the route my previous patches took (/usr/include/qt5 AND
/usr/include/<triplet>/qt5) then much more patching is required than if
we put ALL the headers in /usr/include/<triplet>/qt5. In fact, if all
the headers go to /usr/include/<triplet>/qt5 then upstream doesn't need
patching at all for this. We'd just need this:

 --- a/debian/rules
 +++ b/debian/rules
 @@ -89,7 +89,7 @@ override_dh_auto_configure:
                     -bindir "/usr/lib/$(DEB_HOST_MULTIARCH)/qt5/bin" \
                     -libdir "/usr/lib/$(DEB_HOST_MULTIARCH)" \
                     -docdir "/usr/share/qt5/doc" \
 -                   -headerdir "/usr/include/qt5" \
 +                   -headerdir "/usr/include/$(DEB_HOST_MULTIARCH)/qt5" \
                     -datadir "/usr/share/qt5" \
                     -archdatadir "/usr/lib/$(DEB_HOST_MULTIARCH)/qt5" \
                     -hostdatadir "/usr/share/qt5" \

and some similar tweaks to debian/*.install.

Do you have strong opinions here? Should we make ALL the headers
arch-specific, or just the ones that need to be?

dima


Reply to: