On 09/15/2015 06:31 PM, YunQiang Su wrote: > On Wed, Sep 16, 2015 at 12:27 AM, YunQiang Su <wzss...@gmail.com> wrote: >> On Wed, Sep 16, 2015 at 12:18 AM, Matthias Klose <d...@debian.org> wrote: >>> On 09/13/2015 03:23 PM, YunQiang Su wrote: >>>> Some more fix to debian/rules.d/binary-ada.mk. >>>> >>> >>> >>> - $(cross_shlibdeps) dh_shlibdeps -p$(p_gnsjlj) >>> - $(cross_gencontrol) dh_gencontrol -p$(p_gnsjlj) \ >>> >>> removing the cross_* macros looks wrong, because in your gcc-5-cross patch, >>> you >>> re-add this: >> >> No, it is not wrong. >> With cross_* macros gnat-5-sjlj-<triplet> will have a architecture >> <target> instead >> of <host> >> >> So, don't remove it. >> >>> >>> + sed -i -e "/^Depends:/s/\(libgnat-[0-9]*\) (>= >>> $(DEB_VER_GCC)$(CROSS_EXT))/\1 (>= $(DEB_VER_GCC))/g" \ >>> + tmp/DEBIAN/control ; \ >>> >> >> They are 2 different stories. >> This is for >> gnat-5-<triplet>, which will depends on libgnat-5:<host>. >> If I don't sed it, it will become: >> gnat-5-<triplet>:host depends on libgnat-5:<host> (>= 5.2.1-17cross1) >> >> While in our archive, the version of libgnat-5:<host> is 5.2.1-17. > > If we add a symbol file for libgnat-5, it will resolved, as gnat-5-<triplet> > may > depends on libgnat-5:<host> (>= 5.2.1-16) or lower version, > which is be sed by the previous line with sed.
ok, same as libcc1. However I would like to find a way to add the logic about host vs. target libs in the gcc-5 packaging.