Horton

Re: libpq debug log

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: "'alvherre@alvh.no-ip.org'" <alvherre@alvh.no-ip.org>
Cc: "iwata.aya@fujitsu.com" <iwata.aya@fujitsu.com>, "tsunakawa.takay@fujitsu.com" <tsunakawa.takay@fujitsu.com>, "'Kyotaro Horiguchi'" <horikyota.ntt@gmail.com>, "k.jamison@fujitsu.com" <k.jamison@fujitsu.com>, "pgsql-hackers@lists.postgresql.org" <pgsql-hackers@lists.postgresql.org>
Date: 2021-03-31T20:15:24Z
Lists: pgsql-hackers 
"'alvherre@alvh.no-ip.org'" <alvherre@alvh.no-ip.org> writes:
> So crake failed.  The failure is that it doesn't print the DataRow
> messages.  That's quite odd.  We see this in the trace log:

I think this is a timing problem that's triggered (on some machines)
by force_parallel_mode = regress.  Looking at spurfowl's latest
failure of this type, the postmaster log shows

2021-03-31 14:34:54.982 EDT [18233:15] 001_libpq_pipeline.pl LOG:  execute <unnamed>: SELECT 1.0/g FROM generate_series(3, -1, -1) g
2021-03-31 14:34:54.992 EDT [18234:1] ERROR:  division by zero
2021-03-31 14:34:54.992 EDT [18234:2] STATEMENT:  SELECT 1.0/g FROM generate_series(3, -1, -1) g
2021-03-31 14:34:54.993 EDT [18233:16] 001_libpq_pipeline.pl ERROR:  division by zero
2021-03-31 14:34:54.993 EDT [18233:17] 001_libpq_pipeline.pl STATEMENT:  SELECT 1.0/g FROM generate_series(3, -1, -1) g
2021-03-31 14:34:54.995 EDT [18216:4] LOG:  background worker "parallel worker" (PID 18234) exited with exit code 1
2021-03-31 14:34:54.995 EDT [18233:18] 001_libpq_pipeline.pl LOG:  could not send data to client: Broken pipe
2021-03-31 14:34:54.995 EDT [18233:19] 001_libpq_pipeline.pl FATAL:  connection to client lost

We can see that the division by zero occurred in a parallel worker.
My theory is that in parallel mode, it's uncertain whether the
error will be reported before or after the "preceding" successful
output rows.  So you need to disable parallelism to make this
test case stable.

			regards, tom lane

Commits

  1. Rename PQtraceSetFlags() to PQsetTraceFlags().

  2. Suppress length of Notice/Error msgs in PQtrace regress mode

  3. Strip file names reported in error messages on Windows, too.

  4. Fix setvbuf()-induced crash in libpq_pipeline

  5. libpq_pipeline: Must strdup(optarg) to avoid crash

  6. Remove setvbuf() call from PQtrace()

  7. Initialize conn->Pfdebug to NULL when creating a connection

  8. Disable force_parallel_mode in libpq_pipeline

  9. libpq_pipeline: add PQtrace() support and tests

  10. Improve PQtrace() output format

  11. Re-simplify management of inStart in pqParseInput3's subroutines.