Horton

Thread

  1. Regression on 7.1.2 fails 17/76 - is it up to date?

    Allan Engelhardt <allane@cybaea.com> — 2001-07-02T22:31:57Z

    This *is* supposed to be the forum for naive questions, so here goes: Are the regression tests still valid?

I'm asking, because my installation reports: "17 of 76 tests failed, 1 of these failures ignored."  This seems an awful lot for a recent download.....  It is a known problem, or do I need to investigate further?

There is a bug report relating to PostgreSQL 7.0.3 which seems similar (http://www.ca.postgresql.org/bugs/bugs.php?4~236), but surely this is fixed by now.....?  And if not: any thoughts on whether it is the database or the regression test that is the problem?


Allan.



System info:

RedHat Linux 7.1 on i686 SMP.

bash-2.04$ uname -mrsp
Linux 2.4.2-2 i686 unknown

bash-2.04$ rpm -qi postgresql
Name        : postgresql                   Relocations: (not relocateable)
Version     : 7.1.2                             Vendor: (none)
Release     : 4PGDG                         Build Date: Wed 13 Jun 2001 10:18:46 PM BST
Install date: Fri 29 Jun 2001 08:43:51 PM BST      Build Host: lowen.wgcr.org
Group       : Applications/Databases        Source RPM: postgresql-7.1.2-4PGDG.src.rpm
Size        : 4426307                          License: BSD
URL         : http://www.postgresql.org/
Summary     : PostgreSQL client programs and libraries.
[...]

bash-2.04$ make installcheck | grep -i fail
test char                 ... FAILED
test varchar              ... FAILED
test int8                 ... FAILED
test numeric              ... FAILED
test abstime              ... FAILED
test copy                 ... FAILED
test select               ... FAILED
test select_distinct      ... FAILED
test select_distinct_on   ... FAILED
test select_implicit      ... FAILED
test select_having        ... FAILED
test aggregates           ... FAILED
test random               ... failed (ignored)
test misc                 ... FAILED
test select_views         ... FAILED
test portals_p2           ... FAILED
test limit                ... FAILED

  2. Re: Regression on 7.1.2 fails 17/76 - is it up to date?

    ghaverla@freenet.edmonton.ab.ca — 2001-07-03T01:53:02Z

    On Mon, 2 Jul 2001, Allan Engelhardt wrote:

> This *is* supposed to be the forum for naive questions, so here goes: Are the regression tests still valid?
> 
> I'm asking, because my installation reports: "17 of 76 tests failed, 1 of these failures ignored."  This seems an awful lot for a recent download.....  It is a known problem, or do I need to investigate further?

I've run regression tests before (not this time) and seen
failures.  Usually, this comes from slight differences in
how the various machines which can run PostGreSQL do floating
point arithmetic.  If the baselines for the tests is done on
a Alpha based machine, and you do tests on an Intel based
machine, there will be slight differences from that.  Whether
that causes all of the errors, I don't know.  But you should
be able to look through the testing and see if the answers
are usually close.  There are also some "obvious" things
which cause differences, but off the top of my head I can't
remember what these were.  This is going back a version or 
two when I last compiled PostGreSQL for Solaris 2.5.1 on
a Sparc.  Just my $0.02.

Gord

Matter Realisations     http://www.materialisations.com/
Gordon Haverland, B.Sc. M.Eng. President
101  9504 182 St. NW    Edmonton, AB, CA  T5T 3A7
780/481-8019            ghaverla @ freenet.edmonton.ab.ca
780/993-1274 (cell)

  3. Re: Regression on 7.1.2 fails 17/76 - is it up to date?

    Tom Lane <tgl@sss.pgh.pa.us> — 2001-07-03T01:56:48Z

    Allan Engelhardt <allane@cybaea.com> writes:
> I'm asking, because my installation reports: "17 of 76 tests failed, 1
> of these failures ignored."  This seems an awful lot for a recent
> download..

Your installation seems to be *seriously* broken.  Don't even consider
using it till you find out what the problem is :-(.  Looking at the
detailed regression diffs would be a good start; also read
http://www.ca.postgresql.org/users-lounge/docs/7.1/postgres/regress.html

As of now I believe the abstime test should be expected to show about
two lines of differences, because it had an ill-considered comparison
between 'current' and '30 June 2001' :-(.  Otherwise you should not see
any differences that you can't explain away as one of the phenomena
mentioned on the aforesaid webpage.

			regards, tom lane

  4. Re: Regression on 7.1.2 fails 17/76 - is it up to date?

    Allan Engelhardt <allane@cybaea.com> — 2001-07-03T19:54:45Z

    Tom Lane wrote:

> Allan Engelhardt <allane@cybaea.com> writes:
> > I'm asking, because my installation reports: "17 of 76 tests failed, 1
> > of these failures ignored."  This seems an awful lot for a recent
> > download..
>
> Your installation seems to be *seriously* broken.  Don't even consider

I've had a more careful look, and most (though not all) of my issues seem to stem from one of two problems:

1. My system seems to have an "interesting" definition of character comparisons (i.e. it says 'A' > 'a')
2. The COPY statements do not work because the path is wrong for the RPM build.

My notes so far are below.  Thanks for all the help so far -  I guess I still need soem ideas on the collating sequence.  Couldn't find anything in the manuals.

Allan.




CHAR

1. Fails because

  test=> SELECT 'A' < 'a';
   ?column?
  ----------
   f
  (1 row)

where it expects t.

  test=> \encoding
  SQL_ASCII

The only local environment variable is LANG=en_US, which I don't think
will cause this problem?

VARCHAR

1. This seems to be the same problem as above....

INT8

1. The regression test for int8 fails on my system because it does not
expect the currency symbol in

  test=> select to_char(123, 'L999');
   to_char
  ---------
   $ 123
  (1 row)

I'm not sure what the problem is.  The server is started with
LANG="en_US" and all other locale parameters unset (AFAICT).  This is
the default from the RPM startup script, so I don't know what the
problem is.

NUMERIC

Same problem as int8

ABSTIME

Fails because the date 'Jun 30, 2001' is hard-coded.... :-)

COPY

Fails because of invalid path: assumes
/usr/lib/pgsql/test/regress/foo when path is ./regress/foo (or
/usr/src/redhat/BUILD/postgresql-whatever/src/test/regress/foo)

SELECT_DISTINCT_ON

1. Hmm, lots of "this will fail..." comments in the code, and it does.

=> XXX Further investigation.

SELECT_IMPLICIT

Simply an issue of the character colating sequence -- see above.

SELECT_HAVING

Simply an issue of the character colating sequence -- see above.

AGGREGATES

1. The onek table is not populated, probably becayse of the COPY
problems earlier??

=> XXX Check

RANDOM

Random always fails because onek table is not populated; see above.

MISC

1. Fails because COPY operation fals because of invalid path....see
   above

2. Other problems look like a feature change "the next two queries
   demonstrate how functions generate bogus duplicates"...?

==> XXX Check.

VIEWS

1. Have to look carefully - looks the same.  Length of line???

PORTALS

1. Guessing that the tenk1 and tenk2 tables are not set up correctly
   because of COPY problems....
   => XXX Check