Обсуждение: Intermittent Regression test failure in create_misc

Robert Hentosh (hentosh@io.com) reports a bug with a severity of 1
The lower the number the more severe it is.

Short Description
Intermittent Regression test failure in create_misc

Long Description
After fetching the latest CVS yesterday. I ran the regression test several times in a row. Actually around 10x... in
thattime it failed two times the same way.  I noticed that locking code was changed in CVS after the tag of BETA3 so I
fetchedfrom CVS the REL7_1_BETA3 tag.  This one took around 12 tries but finally failed the same way. 

I am running on a freshly installed OpenBSD 2.8 on a Petium III with 128M real mem and 256M swap. I didn't make any
modificationsto ulimits or other system settings other than adding bison and enabling soft updates (The test failed
beforeand after enabling soft updates). 

Postgresql was compiled with ./configure --enable-syslog after modifying "#define BLCKSZ 16384" in
src/include/config.h.in

Here is the begining of regression.diffs The errors that follow the first one seem to all be related to the fact that
thisone failed and the that onek and onek2 do not exist. 

----------------------------
*** ./expected/create_misc.out  Sat Feb 24 23:14:39 2001
--- ./results/create_misc.out   Mon Feb 26 15:37:15 2001
***************
*** 6,11 ****
--- 6,12 ----
  --
  INSERT INTO tenk2 VALUES (tenk1.*);
  SELECT * INTO TABLE onek2 FROM onek;
+ ERROR:  Relation 'onek' does not exist
  INSERT INTO fast_emp4000 VALUES (slow_emp4000.*);
  SELECT *
     INTO TABLE Bprime

======================================================================

*** ./expected/select_into.out  Sat Feb 24 23:14:48 2001
--- ./results/select_into.out   Mon Feb 26 15:37:48 2001
***************
*** 10,13 ****
--- 10,15 ----
     INTO TABLE tmp1
     FROM onek2
     WHERE onek2.unique1 < 2;
+ ERROR:  Relation 'onek2' does not exist
  DROP TABLE tmp1;
+ ERROR:  table "tmp1" does not exist

======================================================================

*** ./expected/misc.out Sat Feb 24 23:31:36 2001
--- ./results/misc.out  Mon Feb 26 15:38:00 2001
***************
*** 30,35 ****
--- 30,36 ----
     SET stringu1 = reverse_name(onek2.stringu1)
     WHERE onek2.stringu1 = 'JCAAAA' and
          onek2.stringu1 = tmp.stringu1;
+ ERROR:  Relation 'onek2' does not exist
----------------------------


Sample Code


No file was uploaded with this report

Robert Hentosh (hentosh@io.com) writes:
> Postgresql was compiled with ./configure --enable-syslog after
> modifying "#define BLCKSZ 16384" in src/include/config.h.in

Hm.  I wonder if that is triggering the problem, or if it's something
platform-dependent?  Could you rebuild with standard blocksize and
see if you still see the intermittent failures?

Also, are you running the parallel or serial regress tests ("make check"
or "make installcheck")?

            regards, tom lane

On Mon, Feb 26, 2001 at 10:34:05PM -0500, Tom Lane wrote:
> Robert Hentosh (hentosh@io.com) writes:
> > Postgresql was compiled with ./configure --enable-syslog after
> > modifying "#define BLCKSZ 16384" in src/include/config.h.in
>
> Hm.  I wonder if that is triggering the problem, or if it's something
> platform-dependent?  Could you rebuild with standard blocksize and
> see if you still see the intermittent failures?
>
> Also, are you running the parallel or serial regress tests ("make check"
> or "make installcheck")?
>
>             regards, tom lane

I was running a "gmake check" (the parallel tests).  I recompiled with
just the default BLCKSZ, then my errors didn't happen at the same spot
but were erratic.

After searching on the net I came across mention that the system
limits need to be modified for postgresql to run the regression tests.

http://www.crimelabs.net/postgresql.shtml

Knowing what to search for I found some mention in the documentation
under doc/src/sgml/runtime.sgml. After correcting the per user
resource limits in /etc/login.conf. I now have completed running the
regression test over 15 times successfully with the standard BLCKSZ
(except for the expected random test failure).

This issue can be closed.  My mistake.

Where should I have looked to find out that I was exhausting my
resources? I would like to size the system correctly when I test my
application.

I new to BSD, this is the first time I am using it for a development
project.

Thanks,
Robert Hentosh

Robert Hentosh <hentosh@io.com> writes:
> Where should I have looked to find out that I was exhausting my
> resources? I would like to size the system correctly when I test my
> application.

One would like to think that the postmaster log would've showed
resource-related error messages.  However, the regression test shell
doesn't exactly highlight such messages :-(

            regards, tom lane