On 11/3/20 9:25 AM, Christoph Berg wrote: > Re: Sebastiaan Couwenberg >>> -debian/control: debian/control.in debian/pgversions >>> +include /usr/share/postgresql-common/pgxs_debian_control.mk >> >> Please don't do this, this makes building postgis on stable systems >> impossible. > > Morning, > > I can see your point, but wasn't it already like this before?
No, we explicitly added the templating bits to support this use case. > Calling clean was already triggering the control generation: > > override_dh_auto_clean: debian/control While that also changed d/control outside the chroot, it didn't call `pg_buildext checkcontrol` which fails due to those changes. The clean target inside the chroot regenerates d/control again and that will match with what was committed in git. >> The clean target is executed outside the chroot where an older version >> of postgresql will be installed (or none at all), the changes to the >> control file will prevent git-buildpackage from starting the build. >> >> Requiring the same version of postgresql to be installed outside the >> chroot makes building postgresql related packages a pain in the ass. > > Strictly-speaking, B-Ds need to be present for "clean" to work. > > We can try to move the control generation to a different point in the > build process, but "clean" has worked well for me in the past. (My > development systems are all running testing, so the version skew is > not a problem there.) AFAIK only the clean target is suitable, as configure shouldn't modify d/control. Every package that generates d/control has that file a dependency of the clean target. Making `pg_buildext checkcontrol` optional may be a solution for stable systems. Kind Regards, Bas -- GPG Key ID: 4096R/6750F10AE88D4AF1 Fingerprint: 8182 DE41 7056 408D 6146 50D1 6750 F10A E88D 4AF1 _______________________________________________ Pkg-grass-devel mailing list [email protected] https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/pkg-grass-devel
