Forum OpenACS Q&A: Postgres regression suite and text search

Request notifications

Collapse
Posted by Ken Chakiris on
I installed Postgres v7.0beta3 and ran the regression test.  Almost
all of the tests failed.  I also had to hack at the test script in
order to get it to run.  When I looked at the output, it seems that
most of the reasons why the regression failed is not due to bugs but
rather to some seemingly minor ways in which the output differs from
the expected output.  Should I be concerned about the failure of the
regression tests?

What will it take to get the full text search capability working?

Thanks

Collapse
Posted by Don Baccus on
I think you're probably seeing differences due to the overhauled and much improved implementation of psql.  Can you post an example of what  you're seeing?  I've not run the regression tests since my early pre-beta snapshot.

We should be able to kludge up some sort of full text search capability without too much effort, but it will be slow and I don't know if anyone is working on it at present.  Longer-term, adding a smarter, indexed search capability is something I have a personal interest in.  Probably will be late summer or fall before I can get to  it.  There's a simple implementation of such a scheme in the contrib section of Postgres, and PLS at least has some modules that could be used (stemming routines, etc) even though it isn't supported under Linux.  I haven't really investigated this problem in any detail yet.

Collapse
Posted by Lamar Owen on
As of PostgreSQL 7.0beta3 running on RedHat Linux 6.1 (with locale disabled in the OS by removing /etc/sysconfig/i18n), I only get one failure in the regression test -- geometry.

If locale is being used, alot of the tests fail.  Sometimes installing to a nonstandard directory will cause them to fail.  And the parallel tests are pretty finicky.

E-mail me your regression output and diffs (regression.diffs in the regression test dir), and maybe I can help.  Knowing what OS and platform would be helpful.

Collapse
Posted by Ken Chakiris on
I used run_check.sh and got much better results, although there were still a lot of failures.  I have mailed copys of the diffs to both Don and Lamar.
Collapse
Posted by Ken Chakiris on
I have gotten all of tests working except for the trigger and the geometry one.  It seems that I did not compile all of the files needed and some of *.sql files needed to be copied from the ../../../regress/input/ directory.