use binary mode on syslog pipe on windows to avoid upsetting chunking protocol - Mailing list pgsql-patches

From Andrew Dunstan
Subject use binary mode on syslog pipe on windows to avoid upsetting chunking protocol
Date
Msg-id 46ADDC82.1090502@dunslane.net
Whole thread Raw
Responses Re: use binary mode on syslog pipe on windows to avoid upsetting chunking protocol
List pgsql-patches
This small patch makes the syslog pipe use binary mode on Windows so
that CRLF translation (and possibly other oddities) don't upset the pipe
chunking protocol. It preserves text mode for the redirected syslog
file, as recently discussed on -hackers, so there should be no visible
change.

If there is no objection I will apply this and backport it to 8.0 shortly.

cheers

andrew
Index: src/backend/postmaster/postmaster.c
===================================================================
RCS file: /projects/cvsroot/pgsql/src/backend/postmaster/postmaster.c,v
retrieving revision 1.535
diff -c -r1.535 postmaster.c
*** src/backend/postmaster/postmaster.c    24 Jul 2007 04:54:09 -0000    1.535
--- src/backend/postmaster/postmaster.c    30 Jul 2007 12:30:28 -0000
***************
*** 3385,3390 ****
--- 3385,3394 ----

      MyProcPid = getpid();        /* reset MyProcPid */

+ #ifdef WIN32
+     _setmode(fileno(stderr),_O_BINARY);
+ #endif
+
      /* Lose the postmaster's on-exit routines (really a no-op) */
      on_exit_reset();

Index: src/backend/postmaster/syslogger.c
===================================================================
RCS file: /projects/cvsroot/pgsql/src/backend/postmaster/syslogger.c,v
retrieving revision 1.33
diff -c -r1.33 syslogger.c
*** src/backend/postmaster/syslogger.c    19 Jul 2007 19:13:43 -0000    1.33
--- src/backend/postmaster/syslogger.c    30 Jul 2007 12:30:28 -0000
***************
*** 162,167 ****
--- 162,171 ----

      MyProcPid = getpid();        /* reset MyProcPid */

+ #ifdef WIN32
+     _setmode(_fileno(stderr),_O_TEXT);
+ #endif
+
  #ifdef EXEC_BACKEND
      syslogger_parseArgs(argc, argv);
  #endif   /* EXEC_BACKEND */
***************
*** 533,544 ****

                  fflush(stderr);
                  fd = _open_osfhandle((long) syslogPipe[1],
!                                      _O_APPEND | _O_TEXT);
                  if (dup2(fd, _fileno(stderr)) < 0)
                      ereport(FATAL,
                              (errcode_for_file_access(),
                               errmsg("could not redirect stderr: %m")));
                  close(fd);
                  /* Now we are done with the write end of the pipe. */
                  CloseHandle(syslogPipe[1]);
                  syslogPipe[1] = 0;
--- 537,549 ----

                  fflush(stderr);
                  fd = _open_osfhandle((long) syslogPipe[1],
!                                      _O_APPEND | _O_BINARY);
                  if (dup2(fd, _fileno(stderr)) < 0)
                      ereport(FATAL,
                              (errcode_for_file_access(),
                               errmsg("could not redirect stderr: %m")));
                  close(fd);
+                 _setmode(_fileno(stderr),_O_BINARY);
                  /* Now we are done with the write end of the pipe. */
                  CloseHandle(syslogPipe[1]);
                  syslogPipe[1] = 0;
***************
*** 626,632 ****
      fd = atoi(*argv++);
      if (fd != 0)
      {
!         fd = _open_osfhandle(fd, _O_APPEND);
          if (fd > 0)
          {
              syslogFile = fdopen(fd, "a");
--- 631,637 ----
      fd = atoi(*argv++);
      if (fd != 0)
      {
!         fd = _open_osfhandle(fd, _O_APPEND | _O_TEXT);
          if (fd > 0)
          {
              syslogFile = fdopen(fd, "a");
***************
*** 990,995 ****
--- 995,1001 ----

      /* On Windows, need to interlock against data-transfer thread */
  #ifdef WIN32
+     _setmode(_fileno(fh), _O_TEXT); /* use CRLF line endings on Windows */
      EnterCriticalSection(&sysfileSection);
  #endif
      fclose(syslogFile);

pgsql-patches by date:

Previous
From: Gregory Stark
Date:
Subject: Re: Repair cosmetic damage (done by pg_indent?)
Next
From: Gregory Stark
Date:
Subject: Export user visible function to make use of convert_to_scalar