[Debian GNUstep maintainers] Misc issues with gnustep-base

[Yavor Doganov]

Eric Heintzmann wrote:
> I vote for the "gencontrol" script.

IMHO this is the worse solution, because the variables are defined in
debian/rules and are not immediately available to shell scripts.  Here
is an example script (sorry for pasting it inline but the mailing list
is configured to reject attachments):

,----
#!/bin/sh
# Generate debian/control from debian/templates/control.m4.
# The script must be invoked from the top source directory without
# arguments.

set -e

# Read variables defined in debian/rules.
v_objc=${1:-$(debian/rules -qp | awk -F' := ' '$1=="v_objc" {print $2}')}
v_make=${1:-$(debian/rules -qp | awk -F' := ' '$1=="v_make" {print $2}')}
sov_base=${1:-$(debian/rules -qp | awk -F' := ' '$1=="sov_base" {print $2}')}

m4 -DV_OBJC=$v_objc -DV_MAKE=$v_make -DSOV_BASE=$sov_base \
   debian/templates/control.m4 > debian/control
`----

In this case, the other change in debian/rules would be to define the
variables with the "export" directive so that they propagate to the
shell when genscript is invoked within debian/rules.

But the problem is much more serious than the current inconvenience
with the manual generation of debian/control.  The previous version of
the rules was convoluted, but it ensured that debian/control is always
up-to-date.  I just bumbed debhelper to 11 in control.m4 and the
package was happily built.  This is completely unacceptable, it should
have choked with unsatisfied build-dependencies.  Therefore I suggest
to override dh_testdir in order to ensure that any changes are taken
into account before building the package.  WDYT?