[Printing-architecture] CUPS "make test" dependent on cups-filters

[Till Kamppeter <till.kamppeter@gmail.com>]

Hi,

I want to report a problem of CUPS 1.6.4 and 1.7rc1 which should get
fixed in 1.6.5 and 1.7.0 final.

The problem is the build test procedure triggered by running "make test"
(or "make check"). It has the following problems:

1. It depends on the installed cups-filters package. Especially if I add
or remove filters the test in the cups package needs to get adapted.
Also some tests do things requiring the presence of cups-filters, like
printing an image file on a virtual PostScript printer. A build test of
a package should only depend on the content of the package itself. If an
IPP conformance tests is shipped with CUPS it should be separate, for
example an "ipptest" command line utility, usable at run time.

2. The tests fail if cups-filters is in its default configuration, which
has cost factors set in a way that if the input is PostScript and the
printer is PostScript, the pstops filter (part of CUPS) is called
instead of the awkward detour pstopdf->pdftopdf->pdftops. If the cost
factors in cups-filters are changed so that pstopdf->pdftopdf->pdftops
the tests succeed in this case.

See the following Debian bug report about which tests fail:

http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=712237#20

a complete log is attached to the bug report.

What needs to get fixed is either pstops or the test suite. Letting the
cost factors for the filter chain doing an awkward detour which makes
many jobs fail is not a real solution.

   Till