I looked in the code of t/20createdrop.t and t/lib.pl and found that in
lib.pl a lookup in the environment for DBI_DSN, DBI_PASS and DBI_USER is
made. If they are found these settings are used for further testing (for
getting data etc.) - else files for the use as tables are created
(lip.pl lines 28-34). If I unset the DBI_* variables, DBD::Adabas is not
used and all DBD::CSV-tests are passing - and finally 'make install'
succeeds.
May be this DBI_* usage should be removed from lib.pl?
For the DBD::Adabas problem I will open a new thread.
Thanks for all the comments :-)
Robert
Jeff Zucker schrieb:
Robert Roggenbuck wrote:
The DBD::Adabas-error comes during the tests t/20createdrop,
t/30insertfetch, t/40bindparam, and then I stopped going though the
others. The message is exactly the same in every test. If these are
Jeff Zucker's private tests, there is something wrong with the package...
The Adabas and My/Msql stuff in the test directory is left over from
Jochen Wiedmann's original version of the tests which were meant to be a
blueprint for other kinds of tests. I suppose I should clean them out
of the distro, but I've never heard of them being triggered during a
make test of DBD::CSV and there are literally hundreds of CPAN tester
pass reports that successfully ignored the Adabas stuff over the past
years.
So the DBD::CSV problem turns to an DBD::Adabas issue...
But why do I need another DB to test DBD::CSV? This seems to me not
necessary.
No, you absolutely should not need DBD::Adabas to pass the DBD::CSV tests.
