Random regression failures

Bruce Momjian <pgman@candle.pha.pa.us> writes:

I am seeing random parallel regression failures. I would say 20% of
the time I see a failure on the regression test. It has happened
several times to me in the past few days. Regression diff attached.

A stack trace and/or postmaster log report of the crashing process would
be considerably more useful (and less bulky) than this.

regards, tom lane

Bruce Momjian <pgman@candle.pha.pa.us> writes:

I am seeing random parallel regression failures. I would say 20% of
the time I see a failure on the regression test.

FWIW, my other machine has been running repeated parallel regress tests
for several hours. It's now up to 130 completed cycles with no failures
(except one that I deliberately induced to make sure the shell script
would notice...) This is a clean build from a CVS pull at about 14:30
EDT today. All parameters default except for
configure --enable-debug --enable-cassert

regards, tom lane

I ran some loops myself and couldn't reproduce it anymore. I will wait
to see if it happens again.

Sorry.

---------------------------------------------------------------------------

Tom Lane wrote:

Bruce Momjian <pgman@candle.pha.pa.us> writes:

I am seeing random parallel regression failures. I would say 20% of
the time I see a failure on the regression test.

FWIW, my other machine has been running repeated parallel regress tests
for several hours. It's now up to 130 completed cycles with no failures
(except one that I deliberately induced to make sure the shell script
would notice...) This is a clean build from a CVS pull at about 14:30
EDT today. All parameters default except for
configure --enable-debug --enable-cassert

regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 4: Don't 'kill -9' the postmaster

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Bruce Momjian <pgman@candle.pha.pa.us> writes:

I ran some loops myself and couldn't reproduce it anymore. I will wait
to see if it happens again.

Hm. I let mine go for about 400 "make check" cycles and didn't see
anything. Doesn't prove the 401st wouldn't have failed though :-(

I'm wondering whether you may have seen a bug that was fixed in the
recent nested-xact mopup patch. Do you have time to pull CVS from
say Friday or Saturday and see if the problem comes back?

regards, tom lane