new SCO installation FAQ - Mailing list pgsql-hackers

Attached is an installation FAQ for SCO UnixWare and OpenServer related
issues.  It should be placed in the doc/ directory along with the other
platform-specific FAQs.  In addition, it would be nice to have it on the
"Documentation and FAQs" web page, in the "Platform-specific FAQs" list,
along with Linux, Irix, and HP-UX.

Andrew Merrill
The Computer Classroom, Inc., a SCO Authorized Education Center

=======================================================
Frequently Asked Questions (FAQ) for PostgreSQL  V6.5
SCO UnixWare and OpenServer Specific
TO BE READ IN CONJUNCTION WITH THE NORMAL FAQ
=======================================================
last updated:           Tue May 11 12:00:00 PDT 1999

current maintainer:     Andrew Merrill (andrew@compclass.com)
original author:        Andrew Merrill (andrew@compclass.com)


PostgreSQL 6.5 can be built on SCO UnixWare 7 and SCO OpenServer 5.
On OpenServer, you can use either the OpenServer Development Kit or
the Universal Development Kit.

However, some tweaking may be needed, as described below.

Topics:
*) Skunkware
*) GNU Make
*) C++ and libpq++
*) Readline
*) Using the UDK on OpenServer
*) Shared Memory and SHMMAX
*) Java and JDBC
*) Reading the PostgreSQL man pages on UnixWare


***************************************************************************
*) Skunkware

You should locate your copy of the SCO Skunkware CD.  The Skunkware CD
is included with UnixWare 7 and current versions of OpenServer 5.
If you do not have this CD, the software on it
is available via anonymous ftp from ftp.sco.com/skunkware.

For UnixWare 7.1, this CD is now labeled "Open License Software Supplement".

Skunkware has different versions for UnixWare and
OpenServer.  Make sure you install the correct version for your
operating system, except as noted below.


***************************************************************************
*) GNU Make

You need to use the GNU make program, which is on the Skunkware CD.
By default, it installs as /usr/local/bin/make.  To avoid confusion
with the SCO make program, you may want to rename GNU make to gmake.


***************************************************************************
*) C++ and libpq++

I have not been able to build libpq++, the PostgreSQL C++ interface, with
the UnixWare or OpenServer C++ compilers.  By default, building PostgreSQL
also builds the libpq++ interface.  When that fails, it causes the entire
build of PostgreSQL to fail.

This is the problem if you see the following error message:

"pgenv.cc", line 47: error: no default constructor exists for class "string"

If you have this problem, you can disable building of libpq++ with the
following configure option:

configure --without-CXX


***************************************************************************
*) Readline

If you install the readline library, then psql (the PostgreSQL command
line SQL interpreter) remembers each command you type, and allows
you to use arrow keys to recall and edit previous commands.  This is
very helpful, and is strongly recommended.  The readline library is
on the Skunkware CD.

The readline library is not included on the UnixWare 7.1 Skunkware CD.  If
you have the UnixWare 7.0.0 or 7.0.1 Skunkware CDs, you can install it
from there.  Otherwise, try ftp.sco.com/skunkware.

By default, readline installs into /usr/local/lib and /usr/local/include.
However, the PostgreSQL configure program will not find it there without
help.  If you installed readline, then use the following options to configure:

configure --with-libs=/usr/local/lib --with-includes=/usr/local/include

Putting this together with the no-C++ option above yields:

configure --with-libs=/usr/local/lib --with-includes=/usr/local/include --without-CXX


***************************************************************************
*) Using the UDK on OpenServer

If you are using the new Universal Development Kit (UDK) compiler on
OpenServer, you need to use different arguments to the configure program.

First, you need to specify the "unixware" template instead of the default.

Second, you need to specify the locations of the UDK libraries.

Putting these together:

configure --with-template=unixware --with-libs=/udk/usr/lib --with-includes=/udk/usr/include

Putting these together with the no-C++ and readline options from above:

./configure --with-template=unixware --with-libs="/udk/usr/lib /usr/local/lib" --with-includes="/udk/usr/include
/usr/local/include"--without-CXX 


***************************************************************************
*) Shared Memory and SHMMAX

PostgreSQL supports multiple backend daemons running at once.  A block
of shared memory is used by the backend processes.  A larger block
of shared memory allows PostgreSQL to run faster and support more
complicated queries.

By default, UnixWare 7 and OpenServer are confiugured to support shared memory
blocks that are no larger than 524288 bytes, or 512K.  By default, PostgreSQL
tries to allocate a shared memory block that is larger than this.  If
you don't do anything, this allocation will fail, and the postmaster
daemon will not be able to run.

The error message looks like this (the numbers may be different):

IpcMemoryCreate: shmget failed (Invalid argument) key=5432001, size=831176, permission=600
FATAL 1:  ShmemCreate: cannot create region

You have two choices: tell PostgreSQL to allocate a smaller shared memory
block, or tell Unix to allow larger shared memory blocks.  The latter
is the preferred solution, but it requires a kernel tunable change and a
reboot to implement.

To configure the size of the PostgreSQL shared memory block, use the -B
option to the postmaster command, which configures the number of buffers
used by PostgresSQL.  (The shared memory block consists of these buffers
and around 300K of other stuff.)  Each buffer uses 8K, and by default
there are 64 buffers, or 64*8*1024 = 524288 bytes (plus the ~300K of other
stuff).

To use PostgreSQL without doing any kernel tuning, use a -B value of
about 24.  This would take up 24*8*1024 = 196608 bytes, plus ~300K
of other stuff, yields about 500000, which will fit in under the
default 512K limit.

Example: postmaster -B 24

The recommended option is to instead raise the kernel tunable SHMMAX,
which controls the size of the largest allowed shared memory block.

*** Tuning SHMMAX on UnixWare ***

To display the current value of SHMMAX, run:
/etc/conf/bin/idtune -g SHMMAX
which displays the current, default, minimum, and maximum values, in bytes.

To set a new value for SHMMAX, run:
/etc/conf/bin/idtune SHMMAX value
where value is the new value you want to use (in bytes).

After setting SHMMAX, rebuild the kernel and reboot.
To rebuild the kernel:
/etc/conf/bin/idbuild -B

*** Tuning SHMMAX on OpenServer ***

First, cd to /etc/conf/cf.d.

To display the current value of SHMMAX, in bytes, run:
./configure -y SHMMAX

To set a new value for SHMMAX, run:
./configure SHMMAX=value
where value is the new value you want to use (in bytes).

After setting SHMMAX, rebuild the kernel and reboot.
To rebuild the kernel:
./link_unix


***************************************************************************
*) Java and JDBC

The JDBC interface will not build on UnixWare or OpenServer without changes.
The JDBC Makefile in src/interfaces/jdbc/Makefile uses the $$( ) construction
to run an external shell command, instead of the older ` ` syntax.
However, the $$( ) syntax does not work on UnixWare or OpenServer.
So, each of the two uses of it must be replaced with backquotes.  You can
search for $$( to locate the two lines that need changing.

In the file src/interfaces/jdbc/Makefile :

change:
    make $$($(JAVA) makeVersion)
to:
    make `$(JAVA) makeVersion`

and change:
    $(JAR) -c0f $@ $$($(FIND) postgresql -name "*.class" -print)
to:
    $(JAR) -c0f $@ `$(FIND) postgresql -name "*.class" -print`

Of course, you also need to have installed Java on your system, and
make sure that /usr/java/bin is in your PATH.

And, remember to use GNU make, as always.


***************************************************************************
*) Reading the PostgreSQL man pages on UnixWare

By default, the PostgreSQL man pages are installed into /usr/local/pgsql/man.
By default, UnixWare does not look there for man pages, so you will not
be able to read them.

You need to make two changes to access the PostgreSQL man pages from UnixWare.

1) You need to modify the MANPATH environment variable.  I use:

MANPATH=/usr/local/pgsql/man:/usr/dt/man:/usr/man:/usr/share/man:scohelp
export MANPATH

2) The man pages for SQL commands are, by default, placed in section l
(normally used for "l"ocal pages).  UnixWare does not support the l section.

The solution I use is to move all these pages from section l to an unused
section, such as section 6.  To accomplish that:

cd /usr/local/pgsql/man
mv manl man6
cd man6
for file in *.l
do
    mv $file `basename $file .l`.6
done

I have not tried using the PostgreSQL man pages on OpenServer.  Volunteers??