Make pscheduler-server RPM fail when postgresql-load does

[mfeit-internet2]

The calls to postgresql-load in the pscheduler-server RPM spec cause the scriptlet to stop on error, but for some internal, RPM-related reason, the whole operation continues.

Discussion:

• https://bugzilla.redhat.com/show_bug.cgi?id=569930
• https://marc.info/?l=rpm-devel&m=131843439324890&w=2 (Whole thread)
• http://rpm5.org/community/rpm-users/0834.html (Dead link)

Try to find a workaround for this.

[mfeit-internet2]

Consider doing a test load of the finished SQL into a temporary database during the build instead of depending on loads failing during installation.

[mfeit-internet2]

Arranged a syntax check during build to keep silly mistakes from getting into production.

Because this requires a running PostgreSQL server, the Makefile will not do the check unless CHECK_SYNTAX is defined. The RPM spec does this; the Debian packagers can set this if/when they're ready to do this. (Since Debian is effectively a port of the CentOS version, the developers should catch most things before they reach Debian.)

[vvidic]

Sound good to make this optional. We can than try to integrate this check in the test part of the package build or in the post-build package tests.

[vvidic]

On the Debian side the package install will fail if the database does not load properly so basic testing should reveal any problems here.

[mfeit-internet2]

Checking syntax during build seems to have prevented these kinds of problems. Closing.