bad dependency expression results in rule firing on subsequent make, and make install

[gdt]

In several makefiles, e.g extensions/postgis/Makefile.in, there is a rule:

sql_minor_upgrade: sql_bits/extension_upgrade_minor.sql
        for OLD_VERSION in $(UPGRADEABLE_VERSIONS_MINOR); do \
          cat $< > sql/$(EXTENSION)--$$OLD_VERSION--$(EXTVERSION).sql; \
        done

which results in the following being executed on a subsequent make, and on make install, because no file sql_minor_upgrade is created.

---- Making all in postgis
gmake[3]: Entering directory '/u0/n0/gdt/NetBSD-current/pkgsrc/wip/postgresql-postgis2/work/postgis-2.1.3/extensions/postgis'
for OLD_VERSION in 2.0.0 2.0.1 2.0.2 2.0.3 2.0.4 2.0.5; do \
          cat sql_bits/extension_upgrade_minor.sql > sql/postgis--$OLD_VERSION--2.1.3.sql; \
done
for OLD_VERSION in 2.1.0rc1 2.1.0rc2 2.1.0rc3 2.1.0 2.1.1 2.1.2; do \
          cat sql_bits/extension_upgrade_patch.sql > sql/postgis--$OLD_VERSION--2.1.3.sql; \
done
gmake[3]: Leaving directory '/u0/n0/gdt/NetBSD-current/pkgsrc/wip/postgresql-postgis2/work/postgis-2.1.3/extensions/postgis'

To fix, probably the phony target should be created by the rule body, so that it doesn't get reexecuted.

[gdt]

Sorry, this was observed on 2.1.3, not trunk.

[gdt]

patch for discussion (works, but only fixes one subdir)

[gdt]

I am still seeing this in 2.1.8.

[robe]

gdt is this still an issue for you in 2.4? I don't think I've every run into issue.

[robe]

Is this still an issue? anyway pushing to 2.2.8

[gdt]

This is fixed in 2.5.0rc1, so I'm just closing it to clean up, without figuring out where it was fixed along the way.