2 PostgreSQL Installation Guide
3 by The PostgreSQL Development Team
5 PostgreSQL is © 1998-9 by the Postgres Global Development Group.
11 Currently Supported Platforms
14 Requirements to Run Postgres
15 Installation Procedure
19 4. Configuration Options
20 Parameters for Configuration (configure)
21 Parameters for Building (make)
23 What are the Benefits?
24 What are the Drawbacks?
25 Kerberos Authentication
35 Multi-Version Concurrency Control
40 Postgres, developed originally in the UC Berkeley
41 Computer Science Department, pioneered many of the
42 object-relational concepts now becoming available in
43 some commercial databases. It provides SQL92/SQL3
44 language support, transaction integrity, and type
45 extensibility. PostgreSQL is a public-domain, open
46 source descendant of this original Berkeley code.
48 Chapter 1. Introduction
50 This installation procedure makes some assumptions
51 about the desired configuration and runtime
52 environment for your system. This may be adequate for
53 many installations, and is almost certainly adequate
54 for a first installation. But you may want to do an
55 initial installation up to the point of unpacking the
56 source tree and installing documentation, and then
57 print or browse the Administrator's Guide.
61 This manual describes version 6.5.1 of Postgres. The
62 Postgres developer community has compiled and tested
63 Postgres on a number of platforms. Check the web site
64 (http://www.postgresql.org/docs/admin/ports.htm) for
65 the latest information.
67 Currently Supported Platforms
69 At the time of publication, the following platforms
72 Table 2-1. Supported Platforms
73 OS Processor Version Reported Remarks
74 AIX 4.3.2 RS6000 v6.5 1999-05-26 (Andreas Zeugswetter
75 (mailto:Andreas.Zeugswetter@telecom.at))
76 BSDI x86 v6.5 1999-05-25 (Bruce Momjian
77 (mailto:maillist@candle.pha.pa.us)
78 FreeBSD x86 v6.5 1999-05-25 (Tatsuo Ishii
79 2.2.x-4.0 (mailto:t-ishii@sra.co.jp),
81 (mailto:scrappy@hub.org))
82 DGUX m88k v6.3 1998-03-01 v6.4 probably OK.
83 5.4R4.11 Needs new maintainer.
85 (mailto:geek+@cmu.edu))
86 Digital Alpha v6.4 1998-10-29 Minor patchable problems
87 Unix 4.0 (Pedro J. Lobo
88 (mailto:pjlobo@euitt.upm.es))
89 HPUX PA-RISC v6.4 1998-10-25 Both 9.0x and 10.20
90 (Tom Lane (mailto:tgl@sss.pgh.pa.us),
91 Stan Brown (mailto:stanb@awod.com))
92 IRIX 6.5 MIPS v6.4 1998-12-29 IRIX 5.x is different
93 (Mark Dalphin (mdalphin@amgen.com))
94 linux Alpha v6.3.2 1998-04-16 Mostly successful. Needs
97 (mailto:rkirkpat@nag.cs.colorado.edu))
98 linux x86 v6.4 1998-10-27 (Thomas Lockhart
99 2.0.x/libc5 (mailto:lockhart@alumni.caltech.edu))
100 linux x86 v6.4 1999-05-24 (Thomas Lockhart
101 2.0.x/glibc2 (mailto:lockhart@alumni.caltech.edu))
102 linux MIPS v6.4 1998-12-16 Cobalt Qube (Tatsuo Ishii
103 2.0.x (mailto:t-ishii@sra.co.jp))
104 linux Sparc v6.4 1998-10-25 (Tom Szybist
105 2.0.x (mailto:szybist@boxhill.com))
106 linuxPPC PPC603e v6.4 1998-10-26 Powerbook 2400c
108 (mailto:t-ishii@sra.co.jp))
109 mklinux PPC750 v6.4 1998-09-16 PowerMac 7600
111 (mailto:t-ishii@sra.co.jp))
112 NetBSD arm32 v6.5 1999-04-14 (Andrew McMurry
113 (mailto:a.mcmurry1@physics.oxford.ac.uk))
114 NetBSD/i3- x86 v6.4 1998-10-25 (Brook Milligan
115 86 1.3.2 (mailto:brook@trillium.NMSU.Edu))
116 NetBSD m68k v6.4.2 1998-12-28 Mac SE/30 (Mr. Mutsuki
117 Nakajima, Tatsuo Ishii
118 (mailto:t-ishii@sra.co.jp))
119 NetBSD- NS32532 v6.4 1998-10-27 small problems
120 current in date/time math (Jon Buller
121 (mailto:jonb@metronet.com))
122 NetBSD/sp- Sparc v6.4 1998-10-27 (Tom I Helbekkmo
123 arc 1.3H (mailto:tih@hamartun.priv.no))
124 NetBSD 1.3 VAX v6.3 1998-03-01 (Tom I Helbekkmo
125 (mailto:tih@hamartun.priv.no))
126 SCO x86 v6.5 1999-05-25 (Andrew Merrill
127 OpenServer 5 (mailto:andrew@compclass.com))
128 SCO x86 v6.5 1999-05-25 (Andrew Merrill
129 UnixWare 7 (mailto:andrew@compclass.com))
130 Solaris x86 v6.4 1998-10-28 (Marc Fournier
131 (mailto:scrappy@hub.org))
132 Solaris Sparc v6.4 1998-10-28 (Tom Szybist
133 2.6-2.7 (mailto:szybist@boxhill.com),
135 (mailto:ridderbusch.pad@sni.de))
136 SunOS Sparc v6.3 1998-03-01 Patches submitted
138 (mailto:t-ishii@sra.co.jp))
139 SVR4 MIPS v6.4 1998-10-28 No 64-bit int compiler
140 support (Frank Ridderbusch
141 (mailto:ridderbusch.pad@sni.de))
142 Windows x86 v6.4 1999-01-06 Client-side libraries
143 or ODBC/JDBC. No server yet.
146 Windows NT x86 v6.5 1999-05-26 Working with the Cygwin
147 library. (Daniel Horak
148 (mailto:Dan.Horak@email.cz))
152 Platforms listed for v6.3.x and v6.4.x should also
153 work with v6.5.1, but we did not receive explicit
154 confirmation of such at the time this list was
157 Note: For Windows NT, the server-side port of
158 Postgres has recently been accomplished. The
159 Cygnus library is required to compile it.
161 Unsupported Platforms
163 There are a few platforms which have been attempted
164 and which have been reported to not work with the
165 standard distribution. Others listed here do not
166 provide sufficient library support for an attempt.
168 Table 2-2. Possibly Incompatible Platforms
169 OS Processor Version Reported Remarks
170 MacOS all v6.3 1998-03-01 Not library compatible;
172 NextStep x86 v6.x 1998-03-01 Client-only support;
173 v1.0.9 worked with patches
175 (mailto:dave@turbocat.de))
176 SVR4 4.4 m88k v6.2.1 1998-03-01 Confirmed
181 (mailto:dlw@seavme.xroads.com))
182 Ultrix MIPS,VAX? v6.x 1998-03-01 No recent reports;
186 Chapter 3. Installation
188 Complete installation instructions for Postgres
191 Before installing Postgres, you may wish to visit
192 www.postgresql.org (http://www.postgresql.org) for up
193 to date information, patches, etc.
194 These installation instructions assume:
195 o Commands are Unix-compatible. See note below.
196 o Defaults are used except where noted.
197 o User postgres is the Postgres superuser.
198 o The source path is /usr/src/pgsql (other paths are
200 o The runtime path is /usr/local/pgsql (other paths
203 Commands were tested on RedHat Linux version 5.2
204 using the tcsh shell. Except where noted, they will
205 probably work on most systems. Commands like ps and
206 tar may vary wildly between platforms on what options
207 you should use. Use common sense before typing in
209 Our Makefiles require GNU make (called ?gmake? in this
210 document). They will not work with non-GNU make
211 programs. If you have GNU make installed under the
212 name ?make? instead of ?gmake?, then you will use the
213 command make instead. That's OK, but you need to have
214 the GNU form of make to succeed with an installation.
216 Requirements to Run Postgres
218 Up to date information on supported platforms is at
219 http://www.postgresql.org/docs/admin/install.htm
220 (http://www.postgresql.org/docs/admin/install.htm).
221 In general, most Unix-compatible platforms with
222 modern libraries should be able to run Postgres.
223 Although the minimum required memory for running
224 Postgres is as little as 8MB, there are noticable
225 improvements in runtimes for the regression tests
226 when expanding memory up to 96MB on a relatively fast
227 dual-processor system running X-Windows. The rule is
228 you can never have too much memory.
229 Check that you have sufficient disk space. You will
230 need about 30 Mbytes for /usr/src/pgsql, about 5
231 Mbytes for /usr/local/pgsql (excluding your database)
232 and 1 Mbyte for an empty database. The database will
233 temporarily grow to about 20 Mbytes during the
234 regression tests. You will also need about 3 Mbytes
235 for the distribution tar file.
236 We therefore recommend that during installation and
237 testing you have well over 20 Mbytes free under
238 /usr/local and another 25 Mbytes free on the disk
239 partition containing your database. Once you delete
240 the source files, tar file and regression database,
241 you will need 2 Mbytes for /usr/local/pgsql, 1 Mbyte
242 for the empty database, plus about five times the
243 space you would require to store your database data
245 To check for disk space, use
252 Installation Procedure
254 Postgres Installation
255 For a fresh install or upgrading from previous
256 releases of Postgres:
257 1. Read any last minute information and platform
258 specific porting notes. There are some platform
259 specific notes at the end of this file for
260 Ultrix4.x, Linux, BSD/OS and NeXT. There are other
261 files in directory /usr/src/pgsql/doc, including
262 files FAQ-Irix and FAQ-Linux. Also look in
263 directory ftp://ftp.postgresql.org/pub. If there
264 is a file called INSTALL in this directory then
265 this file will contain the latest installation
267 Please note that a "tested" platform in the list
268 given earlier simply means that someone went to
269 the effort at some point of making sure that a
270 Postgres distribution would compile and run on
271 this platform without modifying the code. Since
272 the current developers will not have access to all
273 of these platforms, some of them may not compile
274 cleanly and pass the regression tests in the
275 current release due to minor problems. Any such
276 known problems and their solutions will be posted
277 in ftp://ftp.postgresql.org/pub/INSTALL.
278 2. Create the Postgres superuser account (postgres is
279 commonly used) if it does not already exist.
280 The owner of the Postgres files can be any
281 unprivileged user account. It must not be root,
282 bin, or any other account with special access
283 rights, as that would create a security risk.
284 3. Log in to the Postgres superuser account. Most of
285 the remaining steps in the installation will
286 happen in this account.
288 ftp://ftp.postgresql.org/pub/postgresql-v6.5.1.tar.gz
289 from the Internet. Store it in your home
291 5. Some platforms use flex. If your system uses flex
292 then make sure you have a good version. To check,
295 If the flex command is not found then you
296 probably do not need it. If the version is 2.5.2
297 or 2.5.4 or greater then you are okay. If it is
298 2.5.3 or before 2.5.2 then you will have to
299 upgrade flex. You may get it at
300 ftp://prep.ai.mit.edu/pub/gnu/flex-2.5.4.tar.gz.
301 If you need flex and don't have it or have the
302 wrong version, then you will be told so when you
303 attempt to compile the program. Feel free to skip
304 this step if you aren't sure you need it. If you
305 do need it then you will be told to
306 install/upgrade flex when you try to compile
308 You may want to do the entire flex installation
309 from the root account, though that is not
310 absolutely necessary. Assuming that you want the
311 installation to place files in the usual default
312 areas, type the following:
318 ftp> get flex-2.5.4.tar.gz
320 $ gunzip -c flex-2.5.4.tar.gz | tar xvf -
322 $ configure --prefix=/usr
325 # You must be root when typing the next line:
329 This will update files /usr/man/man1/flex.1,
330 /usr/bin/flex, /usr/lib/libfl.a,
331 /usr/include/FlexLexer.h and will add a link
332 /usr/bin/flex++ which points to flex.
333 6. If you are not upgrading an existing system then
334 skip to step 9. If you are upgrading from 6.5, you
335 do not need to dump/reload or initdb. Simply
336 compile the source code, stop the postmaster, do a
337 "make install", and restart the postmaster.
338 If you are upgrading from 6.4.* or earlier,
339 back up your database. For alpha- and
340 beta-level releases, the database format is liable
341 to change, often every few weeks, with no notice
342 besides a quick comment in the HACKERS mailing
343 list. Full releases always require a dump/reload
344 from previous releases. It is therefore a bad idea
347 Tip: Do not use the pg_dumpall script from v6.0
348 or everything will be owned by the Postgres
351 To dump your fairly recent post-v6.0 database
353 $ pg_dumpall > db.out
354 To use the latest pg_dumpall script on your
355 existing older database before upgrading Postgres,
356 pull the most recent version of pg_dumpall from
357 the new distribution:
359 $ gunzip -c postgresql-v6.5.1.tar.gz \
360 | tar xvf - src/bin/pg_dump/pg_dumpall
361 $ chmod a+x src/bin/pg_dump/pg_dumpall
362 $ src/bin/pg_dump/pg_dumpall > db.out
364 If you wish to preserve object id's (oids), then
365 use the -o option when running pg_dumpall.
366 However, unless you have a special reason for
367 doing this (such as using OIDs as keys in tables),
369 If the pg_dumpall command seems to take a long
370 time and you think it might have died, then, from
371 another terminal, type
373 several times to see if the size of the file is
375 Please note that if you are upgrading from a
376 version prior to Postgres95 v1.09 then you must
377 back up your database, install Postgres95 v1.09,
378 restore your database, then back it up again. You
379 should also read the release notes which should
380 cover any release-specific issues.
383 You must make sure that your database is not
384 updated in the middle of your backup. If
385 necessary, bring down postmaster, edit the
387 /usr/local/pgsql/data/pg_hba.conf to allow
388 only you on, then bring postmaster back up.
392 7. If you are upgrading an existing system then kill
394 $ ps -ax | grep postmaster
395 This should list the process numbers for a number
396 of processes. Type the following line, with pid
397 replaced by the process id for process postmaster.
398 (Do not use the id for process "grep postmaster".)
401 to actually stop the process.
403 Tip: On systems which have Postgres started at
404 boot time, there is probably a startup file
405 which will accomplish the same thing. For
406 example, on my Linux system I can type
407 $ /etc/rc.d/init.d/postgres.init stop
410 8. If you are upgrading an existing system then move
411 the old directories out of the way. If you are
412 short of disk space then you may have to back up
413 and delete the directories instead. If you do
414 this, save the old database in the
415 /usr/local/pgsql/data directory tree. At a
417 /usr/local/pgsql/data/pg_hba.conf.
425 If you are not using /usr/local/pgsql/data as
426 your data directory (check to see if environment
427 variable PGDATA is set to something else) then you
428 will also want to move this directory in the same
430 9. Make new source and install directories. The
431 actual paths can be different for your
432 installation but you must be consistent throughout
435 Note: There are two places in this installation
436 procedure where you will have an opportunity to
437 specify installation locations for programs,
438 libraries, documentation, and other files.
439 Usually it is sufficient to specify these at the
440 gmake install stage of installation.
446 $ chown postgres:postgres pgsql
449 $ chown postgres:postgres pgsql
451 10. Unzip and untar the new source file. Type
453 $ gunzip -c ~/postgresql-v6.5.1.tar.gz | tar xvf -
454 11. Configure the source code for your system. It
455 is this step at which you can specify your actual
456 installation path for the build process (see the
457 --prefix option below). Type
458 $ cd /usr/src/pgsql/src
459 $ ./configure [ options ]
460 a. Among other chores, the configure script
461 selects a system-specific "template" file
462 from the files provided in the template
463 subdirectory. If it cannot guess which one to
464 use for your system, it will say so and exit.
465 In that case you'll need to figure out which
466 one to use and run configure again, this time
467 giving the --with-template=TEMPLATE option to
468 make the right file be chosen.
470 Please Report Problems: If your system is not
471 automatically recognized by configure and
472 you have to do this, please send email to
473 scrappy@hub.org (mailto:scrappy@hub.org)
474 with the output of the program
475 ./config.guess. Indicate what the template
478 b. Choose configuration options. Check
479 Configuration Options for details. However,
480 for a plain-vanilla first installation with
481 no extra options like multi-byte character
482 support or locale collation support it may be
483 adequate to have chosen the installation
484 areas and to run configure without extra
485 options specified. The configure script
486 accepts many additional options that you can
487 use if you don't like the default
488 configuration. To see them all, type
490 Some of the more commonly used ones are:
491 --prefix=BASEDIR Selects a different
492 base directory for the
494 Postgres configuration.
497 --with-template=TEMPLATE
499 TEMPLATE - the template
501 to be in the directory
503 look there for proper values.
504 --with-tcl Build interface
505 libraries and programs requiring
507 libpgtcl, pgtclsh, and pgtksh.
508 --with-perl Build the Perl
510 --with-odbc Build the ODBC
512 --enable-hba Enables Host Based
513 Authentication (DEFAULT)
514 --disable-hba Disables Host Based
516 --enable-locale Enables USE_LOCALE
517 --enable-cassert Enables
521 compiler that the configure
526 compiler that the configure
528 or exclude C++ compilation
530 only affects libpq++ at
532 c. Here is the configure script used on a Sparc
533 Solaris 2.5 system with /opt/postgres
534 specified as the installation base directory:
535 $ ./configure --prefix=/opt/postgres \
536 --with-template=sparc_solaris-gcc
538 --enable-hba --disable-locale
540 Tip: Of course, you may type these three
541 lines all on the same line.
543 12. Install the man and HTML documentation. Type
544 $ cd /usr/src/pgsql/doc
546 The documentation is also available in Postscript
547 format. Look for files ending with .ps.gz in the
549 13. Compile the program. Type
550 $ cd /usr/src/pgsql/src
551 $ gmake all >& make.log &
553 The last line displayed will hopefully be
554 All of PostgreSQL is successfully made. Ready to
556 Remember, ?gmake? may be called ?make? on your system.
557 At this point, or earlier if you wish, type
558 control-C to get out of tail. (If you have
559 problems later on you may wish to examine file
560 make.log for warning and error messages.)
562 Note: You will probably find a number of warning
563 messages in make.log. Unless you have problems
564 later on, these messages may be safely ignored.
566 If the compiler fails with a message stating that
567 the flex command cannot be found then install flex
568 as described earlier. Next, change directory back
569 to this directory, type
571 then recompile again.
572 Compiler options, such as optimization and
573 debugging, may be specified on the command line
574 using the COPT variable. For example, typing
575 $ gmake COPT="-g" all >& make.log &
576 would invoke your compiler's -g option in all
577 steps of the build. See src/Makefile.global.in for
579 14. Install the program. Type
580 $ cd /usr/src/pgsql/src
581 $ gmake install >& make.install.log &
582 $ tail -f make.install.log
583 The last line displayed will be
584 gmake[1]: Leaving directory
585 `/usr/src/pgsql/src/man'
586 At this point, or earlier if you wish, type
587 control-C to get out of tail. Remember, ?gmake? may
588 be called ?make? on your system.
589 15. If necessary, tell your system how to find
590 the new shared libraries. You can do one of the
591 following, preferably the first:
592 a. As root, edit file /etc/ld.so.conf. Add a
595 to the file. Then run command /sbin/ldconfig.
596 b. In a bash shell, type
598 LD_LIBRARY_PATH=/usr/local/pgsql/lib
599 c. In a csh shell, type
600 setenv LD_LIBRARY_PATH
602 Please note that the above commands may vary
603 wildly for different operating systems. Check the
604 platform specific notes, such as those for
605 Ultrix4.x or and for non-ELF Linux.
606 If, when you create the database, you get the
608 pg_id: can't load library 'libpq.so'
609 then the above step was necessary. Simply do this
610 step, then try to create the database again.
611 16. If you used the --with-perl option to
612 configure, check the install log to see whether
613 the Perl module was actually installed. If you've
614 followed our advice to make the Postgres files be
615 owned by an unprivileged userid, then the Perl
616 module won't have been installed, for lack of
617 write privileges on the Perl library directories.
618 You can complete its installation, either now or
619 later, by becoming the user that does own the Perl
620 library (often root) (via su) and doing
621 $ cd /usr/src/pgsql/src/interfaces/perl5
625 17. If it has not already been done, then prepare
626 account postgres for using Postgres. Any account
627 that will use Postgres must be similarly prepared.
628 There are several ways to influence the runtime
629 environment of the Postgres server. Refer to the
630 Administrator's Guide for more information.
632 Note: The following instructions are for a
633 bash/sh shell. Adapt accordingly for other
637 a. Add the following lines to your login
638 environment: shell, ~/.bash_profile:
639 PATH=$PATH:/usr/local/pgsql/bin
640 MANPATH=$MANPATH:/usr/local/pgsql/man
641 PGLIB=/usr/local/pgsql/lib
642 PGDATA=/usr/local/pgsql/data
643 export PATH MANPATH PGLIB PGDATA
646 b. Several regression tests could fail if the
647 user's locale collation scheme is different
648 from that of standard C locale.
649 If you configure and compile Postgres with
650 the --enable-locale option then set locale
651 environment to C (or unset all LC_*
652 variables) by putting these additional lines
653 to your login environment before starting
658 export LC_COLLATE LC_CTYPE LC_COLLATE
664 c. Make sure that you have defined these
665 variables before continuing with the
666 remaining steps. The easiest way to do this
668 $ source ~/.bash_profile
671 18. Create the database installation from your
672 Postgres superuser account (typically account
673 postgres). Do not do the following as root! This
674 would be a major security hole. Type
676 19. Set up permissions to access the database
677 system. Do this by editing file
678 /usr/local/pgsql/data/pg_hba.conf. The
679 instructions are included in the file. (If your
680 database is not located in the default location,
681 i.e. if PGDATA is set to point elsewhere, then the
682 location of this file will change accordingly.)
683 This file should be made read only again once you
684 are finished. If you are upgrading from v6.0 or
685 later you can copy file pg_hba.conf from your old
686 database on top of the one in your new database,
687 rather than redoing the file from scratch.
688 20. Briefly test that the backend will start and
689 run by running it from the command line.
690 a. Start the postmaster daemon running in the
693 $ nohup postmaster -i > pgserver.log 2>&1 &
694 b. Create a database by typing
696 c. Connect to the new database:
698 d. And run a sample query:
699 postgres=> SELECT datetime 'now';
702 f. Remove the test database (unless you will
703 want to use it later for other tests):
705 21. Run postmaster in the background from your
706 Postgres superuser account (typically account
707 postgres). Do not run postmaster from the root
709 Usually, you will want to modify your computer so
710 that it will automatically start postmaster
711 whenever it boots. It is not required; the
712 Postgres server can be run successfully from
713 non-privileged accounts without root intervention.
714 Here are some suggestions on how to do this,
715 contributed by various users.
716 Whatever you do, postmaster must be run by the
717 Postgres superuser (postgres?) and not by root.
718 This is why all of the examples below start by
719 switching user (su) to postgres. These commands
720 also take into account the fact that environment
721 variables like PATH and PGDATA may not be set
722 properly. The examples are as follows. Use them
723 with extreme caution.
724 o If you are installing from a non-privileged
725 account and have no root access, then start the
726 postmaster and send it to the background:
728 $ nohup postmaster > regress.log 2>&1 &
729 o Edit file rc.local on NetBSD or file rc2.d on
730 SPARC Solaris 2.5.1 to contain the following
732 su postgres -c "/usr/local/pgsql/bin/postmaster
733 -S -D /usr/local/pgsql/data"
734 o In FreeBSD 2.2-RELEASE edit
735 /usr/local/etc/rc.d/pgsql.sh to contain the
736 following lines and make it chmod 755 and chown
739 [ -x /usr/local/pgsql/bin/postmaster ] && {
741 /usr/local/pgsql/bin/postmaster
742 -D/usr/local/pgsql/data
743 -S -o -F > /usr/local/pgsql/errlog' &
746 You may put the line breaks as shown above. The
747 shell is smart enough to keep parsing beyond
748 end-of-line if there is an expression unfinished.
749 The exec saves one layer of shell under the
750 postmaster process so the parent is init.
751 o In RedHat Linux add a file
752 /etc/rc.d/init.d/postgres.init which is based on
753 the example in contrib/linux/. Then make a
754 softlink to this file from
755 /etc/rc.d/rc5.d/S98postgres.init.
756 o In RedHat Linux edit file /etc/inittab to add the
757 following as a single line:
758 pg:2345:respawn:/bin/su - postgres -c
759 "/usr/local/pgsql/bin/postmaster
760 -D/usr/local/pgsql/data
761 >> /usr/local/pgsql/server.log 2>&1
763 (The author of this example says this example
764 will revive the postmaster if it dies, but he
765 doesn't know if there are other side effects.)
766 22. Run the regression tests. The file
767 /usr/src/pgsql/src/test/regress/README has
768 detailed instructions for running and interpreting
769 the regression tests. A short version follows
772 $ cd /usr/src/pgsql/src/test/regress
775 You do not need to type gmake clean if this
776 is the first time you are running the tests.
777 You should get on the screen (and also
778 written to file ./regress.out) a series of
779 statements stating which tests passed and
780 which tests failed. Please note that it can
781 be normal for some tests to "fail" on some
782 platforms. The script says a test has failed
783 if there is any difference at all between the
784 actual output of the test and the expected
785 output. Thus, tests may "fail" due to minor
786 differences in wording of error messages,
787 small differences in floating-point roundoff,
788 etc, between your system and the regression
789 test reference platform. "Failures" of this
790 type do not indicate a problem with Postgres.
791 The file ./regression.diffs contains the
792 textual differences between the actual test
793 output on your machine and the "expected"
794 output (which is simply what the reference
795 system produced). You should carefully
796 examine each difference listed to see whether
797 it appears to be a significant issue.
799 o For a i686/Linux-ELF platform, no tests
800 failed since this is the v6.5 regression
801 testing reference platform.
802 Even if a test result clearly indicates a
803 real failure, it may be a localized problem
804 that will not affect you. An example is that
805 the int8 test will fail, producing obviously
806 incorrect output, if your machine and C
807 compiler do not provide a 64-bit integer data
808 type (or if they do but configure didn't
809 discover it). This is not something to worry
810 about unless you need to store 64-bit
812 Conclusion? If you do see failures, try to
813 understand the nature of the differences and
814 then decide if those differences will affect
815 your intended use of Postgres. The regression
816 tests are a helpful tool, but they may
817 require some study to be useful.
818 After running the regression tests, type
819 $ destroydb regression
820 $ cd /usr/src/pgsql/src/test/regress
822 to recover the disk space used for the
823 tests. (You may want to save the
824 regression.diffs file in another place before
826 23. If you haven't already done so, this would be
827 a good time to modify your computer to do regular
828 maintainence. The following should be done at
831 Minimal Backup Procedure
832 1. Run the SQL command VACUUM. This will clean
834 2. Back up your system. (You should probably
835 keep the last few backups on hand.) Preferably,
836 no one else should be using the system at the
839 Ideally, the above tasks should be done by a
840 shell script that is run nightly or weekly by
841 cron. Look at the man page for crontab for a
842 starting point on how to do this. (If you do it,
843 please e-mail us a copy of your shell script. We
844 would like to set up our own systems to do this
846 24. If you are upgrading an existing system then
847 reinstall your old database. Type
849 $ psql -e template1 < db.out
850 If your pre-v6.2 database uses either path or
851 polygon geometric data types, then you will need
852 to upgrade any columns containing those types. To
853 do so, type (from within psql)
854 UPDATE FirstTable SET PathCol =
855 UpgradePath(PathCol);
856 UPDATE SecondTable SET PathCol =
857 UpgradePath(PathCol);
860 UpgradePath() checks to see that a path value is
861 consistant with the old syntax, and will not
862 update a column which fails that examination.
863 UpgradePoly() cannot verify that a polygon is in
864 fact from an old syntax, but RevertPoly() is
865 provided to reverse the effects of a mis-applied
867 25. If you are a new user, you may wish to play
868 with Postgres as described below.
869 26. Clean up after yourself. Type
870 $ rm -rf /usr/src/pgsql_6_5
871 $ rm -rf /usr/local/pgsql_6_5
872 # Also delete old database directory tree if it is
874 # /usr/local/pgsql_6_5/data
875 $ rm ~/postgresql-v6.5.1.tar.gz
876 27. You will probably want to print out the
877 documentation. If you have a Postscript printer,
878 or have your machine already set up to accept
879 Postscript files using a print filter, then to
880 print the User's Guide simply type
881 $ cd /usr/local/pgsql/doc
882 $ gunzip user.ps.tz | lpr
883 Here is how you might do it if you have
884 Ghostscript on your system and are writing to a
886 $ alias gshp='gs -sDEVICE=laserjet -r300
889 GS_LIB=/usr/share/ghostscript:/usr/share/ghostscr-
892 $ gshp -sOUTPUTFILE=user.hp user.ps
894 $ lpr -l -s -r manpage.hp
895 28. The Postgres team wants to keep Postgres
896 working on all of the supported platforms. We
897 therefore ask you to let us know if you did or did
898 not get Postgres to work on you system. Please
899 send a mail message to pgsql-ports@postgresql.org
900 (mailto:pgsql-ports@postgresql.org) telling us the
902 o The version of Postgres (v6.5.1, 6.5, beta
904 o Your operating system (i.e. RedHat v5.2 Linux
906 o Your hardware (SPARC, i486, etc.).
907 o Did you compile, install and run the regression
908 tests cleanly? If not, what source code did you
909 change (i.e. patches you applied, changes you
910 made, etc.), what tests failed, etc. It is normal
911 to get many warning when you compile. You do not
912 need to report these.
913 29. Now create, access and manipulate databases
914 as desired. Write client programs to access the
915 database server. In other words, enjoy!
917 Playing with Postgres
919 After Postgres is installed, a database system is
920 created, a postmaster daemon is running, and the
921 regression tests have passed, you'll want to see
922 Postgres do something. That's easy. Invoke the
923 interactive interface to Postgres, psql:
927 (psql has to open a particular database, but at this
928 point the only one that exists is the template1
929 database, which always exists. We will connect to it
930 only long enough to create another one and switch to
932 The response from psql is:
934 Welcome to the POSTGRESQL interactive sql monitor:
935 Please read the file COPYRIGHT for copyright terms
938 type \? for help on slash commands
940 type \g or terminate with semicolon to execute
942 You are currently connected to the database:
947 Create the database foo:
949 template1=> create database foo;
952 (Get in the habit of including those SQL semicolons.
953 Psql won't execute anything until it sees the
954 semicolon or a "\g" and the semicolon is required to
955 delimit multiple statements.)
956 Now connect to the new database:
959 connecting to new database: foo
961 ("slash" commands aren't SQL, so no semicolon. Use \?
962 to see all the slash commands.)
965 foo=> create table bar (i int4, c char(16));
968 Then inspect the new table:
973 +----------------------------------+-----------------
974 ------------------+-------+
977 +----------------------------------+-----------------
978 ------------------+-------+
983 +----------------------------------+-----------------
984 ------------------+-------+
986 And so on. You get the idea.
990 Questions? Bugs? Feedback? First, read the files in
991 directory /usr/src/pgsql/doc/. The FAQ in this
992 directory may be particularly useful.
993 If Postgres failed to compile on your computer then
994 fill out the form in file
995 /usr/src/pgsql/doc/bug.template and mail it to the
996 location indicated at the top of the form.
997 Check on the web site at http://www.postgresql.org
998 For more information on the various support mailing
1003 Check for any platform-specific FAQs in the doc/
1004 directory of the source distribution.
1006 Chapter 4. Configuration Options
1008 Parameters for Configuration (configure)
1010 The full set of parameters available in configure
1011 can be obtained by typing
1013 $ ./configure --help
1017 The following parameters may be of interest to
1020 Directory and file names:
1021 --prefix=PREFIX install
1022 architecture-independent files in PREFIX
1024 --bindir=DIR user executables in DIR
1026 --libdir=DIR object code libraries in
1028 --includedir=DIR C header files in DIR
1030 --mandir=DIR man documentation in DIR
1032 Features and packages:
1033 --disable-FEATURE do not include FEATURE
1034 (same as --enable-FEATURE=no)
1035 --enable-FEATURE[=ARG] include FEATURE [ARG=yes]
1036 --with-PACKAGE[=ARG] use PACKAGE [ARG=yes]
1037 --without-PACKAGE do not use PACKAGE (same as
1039 --enable and --with options recognized:
1040 --with-template=template
1041 use operating system
1043 see template directory
1044 --with-includes=incdir site header files for
1046 --with-libs=incdir also search for libraries
1048 --with-libraries=libdir also search for libraries
1050 --enable-locale enable locale support
1051 --enable-recode enable cyrillic recode
1053 --with-mb=encoding enable multi-byte support
1054 --with-pgport=portnum change default startup port
1055 --with-maxbackends=n set default maximum number of
1057 --with-tcl build Tcl interfaces and
1059 --with-tclconfig=tcldir tclConfig.sh and
1060 tkConfig.sh are in DIR
1061 --with-perl build Perl interface
1062 --with-odbc build ODBC driver package
1063 --with-odbcinst=odbcdir change default directory
1065 --enable-cassert enable assertion checks
1067 --with-CC=compiler use specific C compiler
1068 --with-CXX=compiler use specific C++ compiler
1069 --without-CXX prevent building C++ code
1073 Some systems may have trouble building a specific
1074 feature of Postgres. For example, systems with a
1075 damaged C++ compiler may need to specify
1076 --without-CXX to instruct the build procedure to skip
1077 construction of libpq++.
1079 Parameters for Building (make)
1081 Many installation-related parameters can be set in
1082 the building stage of Postgres installation.
1083 In most cases, these parameters should be placed in
1084 a file, Makefile.custom, intended just for that
1085 purpose. The default distribution does not contain
1086 this optional file, so you will create it using a
1087 text editor of your choice. When upgrading
1088 installations, you can simply copy your old
1089 Makefile.custom to the new installation before doing
1092 make [ variable=value [,...] ]
1096 A few of the many variables which can be specified
1100 Top of the installation tree.
1103 Location of applications and utilities.
1106 Location of object libraries, including shared
1110 Location of include files.
1113 Location of installation-wide psqlODBC (ODBC)
1116 There are other optional parameters which are not as
1117 commonly used. Many of those listed below are
1118 appropriate when doing Postgres server code
1122 Set flags for the C compiler. Should be assigned
1123 with "+=" to retain relevant default parameters.
1126 Set flags for the yacc/bison parser. -v might be
1127 used to help diagnose problems building a new
1128 parser. Should be assigned with "+=" to retain
1129 relevant default parameters.
1132 Enable Tcl interface building.
1135 DocBook HTML style sheets for building the
1136 documentation from scratch. Not used unless you
1137 are developing new documentation from the
1138 DocBook-compatible SGML source documents in
1142 DocBook style sheets for building printed
1143 documentation from scratch. Not used unless you
1144 are developing new documentation from the
1145 DocBook-compatible SGML source documents in
1148 Here is an example Makefile.custom for a PentiumPro
1152 # Thomas Lockhart 1998-03-01
1154 POSTGRESDIR= /opt/postgres/current
1155 CFLAGS+= -m486 # -g -O0
1158 X_LIBS= -L/usr/X11/lib
1163 HSTYLE= /home/tgl/SGML/db118.d/docbook/html
1164 PSTYLE= /home/tgl/SGML/db118.d/docbook/print
1173 Note: Written by Oleg Bartunov. See Oleg's web
1174 page (http://www.sai.msu.su/~megera/postgres/) for
1175 additional information on locale and Russian
1178 While doing a project for a company in Moscow,
1179 Russia, I encountered the problem that postgresql had
1180 no support of national alphabets. After looking for
1181 possible workarounds I decided to develop support of
1182 locale myself. I'm not a C-programer but already had
1183 some experience with locale programming when I work
1184 with perl (debugging) and glimpse. After several days
1185 of digging through the Postgres source tree I made
1186 very minor corections to
1187 src/backend/utils/adt/varlena.c and
1188 src/backend/main/main.c and got what I needed! I did
1189 support only for LC_CTYPE and LC_COLLATE, but later
1190 LC_MONETARY was added by others. I got many messages
1191 from people about this patch so I decided to send it
1192 to developers and (to my surprise) it was
1193 incorporated into the Postgres distribution.
1194 People often complain that locale doesn't work for
1195 them. There are several common mistakes:
1196 o Didn't properly configure postgresql before
1197 compilation. You must run configure with
1198 --enable-locale option to enable locale support.
1199 Didn't setup environment correctly when starting
1200 postmaster. You must define environment variables
1201 LC_CTYPE and LC_COLLATE before running postmaster
1202 because backend gets information about locale from
1203 environment. I use following shell script
1207 export LC_CTYPE=koi8-r
1208 export LC_COLLATE=koi8-r
1209 postmaster -B 1024 -S
1210 -D/usr/local/pgsql/data/ -o '-Fe'
1212 and run it from rc.local as
1213 /bin/su - postgres -c
1214 "/home/postgres/runpostgres"
1217 o Broken locale support in OS (for example, locale
1218 support in libc under Linux several times has
1219 changed and this caused a lot of problems). Latest
1220 perl has also support of locale and if locale is
1221 broken perl -v will complain something like:
1222 8:17[mira]:~/WWW/postgres>setenv LC_CTYPE
1224 8:18[mira]:~/WWW/postgres>perl -v
1225 perl: warning: Setting locale failed.
1226 perl: warning: Please check that your locale
1229 LC_CTYPE = "not_exist",
1231 are supported and installed on your system.
1232 perl: warning: Falling back to the standard
1236 o Wrong location of locale files! Possible locations
1237 include: /usr/lib/locale (Linux, Solaris),
1238 /usr/share/locale (Linux), /usr/lib/nls/loc (DUX
1239 4.0). Check man locale to find the correct
1240 location. Under Linux I did a symbolic link between
1241 /usr/lib/locale and /usr/share/locale to be sure
1242 that the next libc will not break my locale.
1245 What are the Benefits?
1247 You can use ~* and order by operators for strings
1248 contain characters from national alphabets.
1249 Non-english users definitely need that. If you won't
1250 use locale stuff just undefine the USE_LOCALE
1253 What are the Drawbacks?
1255 There is one evident drawback of using locale - its
1256 speed! So, use locale only if you really need it.
1258 Kerberos Authentication
1260 Kerberos is an industry-standard secure
1261 authentication system suitable for distributed
1262 computing over a public network.
1266 The Kerberos authentication system is not
1267 distributed with Postgres. Versions of Kerberos are
1268 typically available as optional software from
1269 operating system vendors. In addition, a source code
1270 distribution may be obtained through MIT Project
1271 Athena (ftp://athena-dist.mit.edu).
1273 Note: You may wish to obtain the MIT version even
1274 if your vendor provides a version, since some
1275 vendor ports have been deliberately crippled or
1276 rendered non-interoperable with the MIT version.
1278 Users located outside the United States of America
1279 and Canada are warned that distribution of the actual
1280 encryption code in Kerberos is restricted by U. S.
1281 Government export regulations.
1282 Inquiries regarding your Kerberos should be directed
1283 to your vendor or MIT Project Athena
1284 (info-kerberos@athena.mit.edu). Note that FAQLs
1285 (Frequently-Asked Questions Lists) are periodically
1286 posted to the Kerberos mailing list
1287 (mailto:kerberos@ATHENA.MIT.EDU) (send mail to
1288 subscribe (mailto:kerberos-request@ATHENA.MIT.EDU)),
1289 and USENET news group (news:comp.protocols.kerberos).
1293 Installation of Kerberos itself is covered in detail
1294 in the Kerberos Installation Notes . Make sure that
1295 the server key file (the srvtab or keytab) is somehow
1296 readable by the Postgres account.
1297 Postgres and its clients can be compiled to use
1298 either Version 4 or Version 5 of the MIT Kerberos
1299 protocols by setting the KRBVERS variable in the file
1300 src/Makefile.global to the appropriate value. You can
1301 also change the location where Postgres expects to
1302 find the associated libraries, header files and its
1303 own server key file.
1304 After compilation is complete, Postgres must be
1305 registered as a Kerberos service. See the Kerberos
1306 Operations Notes and related manual pages for more
1307 details on registering services.
1311 After initial installation, Postgres should operate
1312 in all ways as a normal Kerberos service. For details
1313 on the use of authentication, see the PostgreSQL
1314 User's Guide reference sections for postmaster and
1316 In the Kerberos Version 5 hooks, the following
1317 assumptions are made about user and service naming:
1318 o User principal names (anames) are assumed to
1319 contain the actual Unix/Postgres user name in the
1321 o The Postgres service is assumed to be have two
1322 components, the service name and a hostname,
1323 canonicalized as in Version 4 (i.e., with all
1324 domain suffixes removed).
1328 Table 4-1. Kerberos Parameter Examples
1331 user aoki/HOST=miyu.S2K.Berkeley.EDU@S2K.ORG
1332 host postgres_dbms/ucbvax@S2K.ORG
1336 Support for Version 4 will disappear sometime after
1337 the production release of Version 5 by MIT.
1339 Chapter 5. Release Notes
1343 This release marks a major step in the development
1344 team's mastery of the source code we inherited from
1345 Berkeley. You will see we are now easily adding major
1346 features, thanks to the increasing size and
1347 experience of our world-wide development team.
1348 Here is a brief summary of some of the more
1351 Multi-version concurrency control(MVCC)
1352 This removes our old table-level locking, and
1353 replaces it with a locking system that is superior
1354 to most commercial database systems. In a
1355 traditional system, each row that is modified is
1356 locked until committed, preventing reads by other
1357 users. MVCC uses the natural multi-version nature
1358 of PostgreSQL to allow readers to continue reading
1359 consistent data during writer activity. Writers
1360 continue to use the compact pg_log transaction
1361 system. This is all performed without having to
1362 allocate a lock for every row like traditional
1363 database systems. So, basically, we no longer are
1364 restricted by simple table-level locking; we have
1365 something better than row-level locking.
1368 We now have a true numeric data type, with
1369 user-specified precision.
1372 Temporary tables are guaranteed to have unique
1373 names within a database session, and are destroyed
1377 We now have CASE, INTERSECT, and EXCEPT statement
1378 support. We have new LIMIT/OFFSET, SET TRANSACTION
1379 ISOLATION LEVEL, SELECT ... FOR UPDATE, and an
1380 improved LOCK command.
1383 We continue to speed up PostgreSQL, thanks to the
1384 variety of talents within our team. We have sped
1385 up memory allocation, optimization, table joins,
1386 and row transfer routines.
1389 We continue to expand our port list, this time
1390 including WinNT/ix86 and NetBSD/arm32.
1393 Most interfaces have new versions, and existing
1394 functionality has been improved.
1399 A dump/restore using pg_dump or pg_dumpall is
1400 required for those wishing to migrate data from any
1401 previous release of Postgres.
1402 The new Multi-Version Concurrency Control (MVCC)
1403 features can give somewhat different behaviors in
1404 multi-user environments. Read and understand the
1405 following section to ensure that your existing
1406 applications will give you the behavior you need.
1408 Multi-Version Concurrency Control
1409 Because readers in 6.5 don't lock data, regardless
1410 of transaction isolation level, data read by one
1411 transaction can be overwritten by another. In the
1412 other words, if a row is returned by SELECT it
1413 doesn't mean that this row really exists at the time
1414 it is returned (i.e. sometime after the statement or
1415 transaction began) nor that the row is protected from
1416 deletion or updation by concurrent transactions
1417 before the current transaction does a commit or
1419 To ensure the actual existance of a row and protect
1420 it against concurrent updates one must use SELECT FOR
1421 UPDATE or an appropriate LOCK TABLE statement. This
1422 should be taken into account when porting
1423 applications from previous releases of Postgres and
1425 Keep above in mind if you are using contrib/refint.*
1426 triggers for referential integrity. Additional
1427 technics are required now. One way is to use LOCK
1428 parent_table IN SHARE ROW EXCLUSIVE MODE command if a
1429 transaction is going to update/delete a primary key
1430 and use LOCK parent_table IN SHARE MODE command if a
1431 transaction is going to update/insert a foreign key.
1433 Note: Note that if you run a transaction in
1434 SERIALIZABLE mode then you must execute LOCK
1435 commands above before execution of any DML
1437 (SELECT/INSERT/DELETE/UPDATE/FETCH/COPY_TO) in the
1441 These inconveniences will disappear in the future
1442 when the ability to read dirty (uncommitted) data
1443 (regardless of isolation level) and true referential
1444 integrity will be implemented.
1446 Detailed Change List
1452 Fix text<->float8 and text<->float4 conversion
1454 Fix for creating tables with mixed-case
1456 Change exp()/pow() behavior to generate error on
1457 underflow/overflow(Jan)
1458 Fix bug in pg_dump -z
1459 Memory overrun cleanups(Tatsuo)
1460 Fix for lo_import crash(Tatsuo)
1461 Adjust handling of data type names to suppress double
1463 Use type coersion for matching columns and
1465 Fix deadlock so it only checks once after one second
1467 Fixes for aggregates and PL/pgsql(Hiroshi)
1468 Fix for subquery crash(Vadim)
1469 Fix for libpq function PQfnumber and case-insensitive
1470 names(Bahman Rafatjoo)
1471 Fix for large object write-in-middle, no extra block,
1472 memory consumption(Tatsuo)
1473 Fix for pg_dump -d or -D and quote special
1474 characters in INSERT
1475 Repair serious problems with dynahash(Tom)
1476 Fix INET/CIDR portability problems
1477 Fix problem with selectivity error in ALTER TABLE ADD
1479 Fix executor so mergejoin of different column types
1481 Fix for Alpha OR selectivity bug
1482 Fix OR index selectivity problem(Bruce)
1483 Fix so \d shows proper length for
1484 char()/varchar()(Ryan)
1485 Fix tutorial code(Clark)
1486 Improve destroyuser checking(Oliver)
1487 Fix for Kerberos(Rodney McDuff)
1488 Fix for dropping database while dirty buffers(Bruce)
1489 Fix so sequence nextval() can be
1490 case-sensitive(Bruce)
1492 Drop buffers before destroying database files(Bruce)
1493 Fix case where executor evaluates functions
1495 Allow sequence nextval actions to be
1496 case-sensitive(Bruce)
1497 Fix optimizer indexing not working for negative
1499 Fix for memory leak in executor with fjIsNull
1500 Fix for aggregate memory leaks(Erik Riedel)
1501 Allow username containing a dash GRANT permissions
1502 Cleanup of NULL in inet types
1503 Clean up system table bugs(Tom)
1504 Fix problems of PAGER and \? command(Masaaki Sakaida)
1505 Reduce default multi-segment file size limit to
1507 Fix for dumping of CREATE OPERATOR(Tom)
1508 Fix for backward scanning of cursors(Hiroshi Inoue)
1509 Fix for COPY FROM STDIN when using \i(Tom)
1510 Fix for subselect is compared inside an
1512 Fix handling of error reporting while returning
1514 Fix problems with reference to array types(Tom,Jan)
1515 Prevent UPDATE SET oid(Jan)
1516 Fix pg_dump so -t option can handle case-sensitive
1518 Fixes for GROUP BY in special cases(Tom, Jan)
1519 Fix for memory leak in failed queries(Tom)
1520 DEFAULT now supports mixed-case identifiers(Tom)
1521 Fix for multi-segment uses of DROP/RENAME table,
1526 Add "vacuumdb" utility
1527 Speed up libpq by allocating memory better(Tom)
1528 EXPLAIN all indices used(Tom)
1529 Implement CASE, COALESCE, NULLIF expression(Thomas)
1530 New pg_dump table output format(Constantin)
1531 Add string min()/max() functions(Thomas)
1532 Extend new type coersion techniques to
1534 New moddatetime contrib(Terry)
1535 Update to pgaccess 0.96(Constantin)
1536 Add routines for single-byte "char" type(Thomas)
1537 Improved substr() function(Thomas)
1538 Improved multi-byte handling(Tatsuo)
1539 Multi-version concurrency control/MVCC(Vadim)
1540 New Serialized mode(Vadim)
1541 Fix for tables over 2gigs(Peter)
1542 New SET TRANSACTION ISOLATION LEVEL(Vadim)
1543 New LOCK TABLE IN ... MODE(Vadim)
1544 Update ODBC driver(Byron)
1545 New NUMERIC data type(Jan)
1546 New SELECT FOR UPDATE(Vadim)
1547 Handle "NaN" and "Infinity" for input values(Jan)
1548 Improved date/year handling(Thomas)
1549 Improved handling of backend connections(Magnus)
1550 New options ELOG_TIMESTAMPS and USE_SYSLOG options
1551 for log files(Massimo)
1552 New TCL_ARRAYS option(Massimo)
1553 New INTERSECT and EXCEPT(Stefan)
1554 New pg_index.indisprimary for primary key
1556 New pg_dump option to allow dropping of tables before
1558 Speedup of row output routines(Tom)
1559 New READ COMMITTED isolation level(Vadim)
1560 New TEMP tables/indexes(Bruce)
1561 Prevent sorting if result is already sorted(Jan)
1562 New memory allocation optimization(Jan)
1563 Allow psql to do \p\g(Bruce)
1564 Allow multiple rule actions(Jan)
1565 Added LIMIT/OFFSET functionality(Jan)
1566 Improve optimizer when joining a large number of
1568 New intro to SQL from S. Simkovics' Master's Thesis
1570 New intro to backend processing from S. Simkovics'
1571 Master's Thesis (Stefan)
1572 Improved int8 support(Ryan Bradetich, Thomas, Tom)
1573 New routines to convert between int8 and text/varchar
1575 New bushy plans, where meta-tables are joined(Bruce)
1576 Enable right-hand queries by default(Bruce)
1577 Allow reliable maximum number of backends to be set
1579 (--with-maxbackends and postmaster switch (-N
1581 GEQO default now 10 tables because of optimizer
1583 Allow NULL=Var for MS-SQL portability(Michael, Bruce)
1584 Modify contrib check_primary_key() so either
1585 "automatic" or "dependent"(Anand)
1586 Allow psql \d on a view show query(Ryan)
1587 Speedup for LIKE(Bruce)
1588 Ecpg fixes/features, see
1589 src/interfaces/ecpg/ChangeLog file(Michael)
1590 JDBC fixes/features, see
1591 src/interfaces/jdbc/CHANGELOG(Peter)
1592 Make % operator have precedence like /(Bruce)
1593 Add new postgres -O option to allow system table
1594 structure changes(Bruce)
1595 Update contrib/pginterface/findoidjoins script(Tom)
1596 Major speedup in vacuum of deleted rows with
1598 Allow non-SQL functions to run different versions
1599 based on arguments(Tom)
1600 Add -E option that shows actual queries sent by \dt
1601 and friends(Masaaki Sakaida)
1602 Add version number in startup banners for
1603 psql(Masaaki Sakaida)
1604 New contrib/vacuumlo removes large objects not
1606 New initialization for table sizes so non-vacuumed
1607 tables perform better(Tom)
1608 Improve error messages when a connection is
1610 Support for arrays of char() and varchar()
1612 Overhaul of hash code to increase reliability and
1614 Update to PyGreSQL 2.4(D'Arcy)
1615 Changed debug options so -d4 and -d5 produce
1616 different node displays(Jan)
1617 New pg_options: pretty_plan, pretty_parse,
1618 pretty_rewritten(Jan)
1619 Better optimization statistics for system table
1621 Better handling of non-default block sizes(Massimo)
1622 Improve GEQO optimizer memory consumption(Tom)
1623 UNION now suppports ORDER BY of columns not in target
1625 Major libpq++ improvements(Vince Vielhaber)
1629 Improve port matching(Tom)
1630 Portability fixes for SunOS
1631 Add NT/Win32 backend port and enable dynamic
1632 loading(Magnus and Daniel Horak)
1633 New port to Cobalt Qube(Mips) running Linux(Tatsuo)
1634 Port to NetBSD/m68k(Mr. Mutsuki Nakajima)
1635 Port to NetBSD/sun3(Mr. Mutsuki Nakajima)
1636 Port to NetBSD/macppc(Toshimi Aoki)
1637 Fix for tcl/tk configuration(Vince)
1638 Removed CURRENT keyword for rule queries(Jan)
1639 NT dynamic loading now works(Daniel Horak)
1640 Add ARM32 support(Andrew McMurry)
1641 Better support for HPUX 11 and Unixware
1642 Improve file handling to be more uniform, prevent
1643 file descriptor leak(Tom)
1644 New install commands for plpgsql(Jan)