Re: Syslogger tries to write to /dev/null on Windows

Tom Lane wrote:
> Hmm.  I agree with your proposed change, but it seems to me that there
> is still another mystery here: why does the mistaken open() argument
> lead to a crash?  Per the second comment, this code is supposed to keep
> working even if the open() fails.  If it fails because of that, we have
> robustness problems everywhere not only on Windows --- consider ENFILE
> or some such.

Hmm, good question.

The open() fails and returns a return code as you would expect. But the
dup2() call crashes when passed an invalid file descriptor, I just
tested that with a small test program on Windows.

Googling around, this seems to be because of a feature called Parameter
Validation: http://msdn.microsoft.com/en-us/library/ksazx244.aspx

Following the example there to set a custom Invalid Parameter Handler
Routine makes dup2() not crash on an invalid file handle. We could do
that in PostgreSQL, setting a handler that reports a warning in the log
and continues. Or we could just be more careful when passing parameters
to system functions; this is the first time we hear about this so it
doesn't seem to be a very widespread issue. In this case we should do this:

*** a/src/backend/postmaster/syslogger.c
--- b/src/backend/postmaster/syslogger.c
***************
*** 194,202 ****
           */
          close(fileno(stdout));
          close(fileno(stderr));
!         dup2(fd, fileno(stdout));
!         dup2(fd, fileno(stderr));
!         close(fd);
      }

      /*
--- 194,205 ----
           */
          close(fileno(stdout));
          close(fileno(stderr));
!         if (fd != -1)
!         {
!             dup2(fd, fileno(stdout));
!             dup2(fd, fileno(stderr));
!             close(fd);
!         }
      }

      /*

--
  Heikki Linnakangas
  EnterpriseDB   http://www.enterprisedb.com