Safe and Happy UNIX Hacking with MacOS X

Preface

As I mentioned before, MacOS X version 10.1 lacks package management system similar to RPM or Debian APT which tracks system components in central search-able database (may be Apple will fix this in one of the next versions). This makes quite difficult maintaining UNIX inherited software, which may consist of hundreds and even thousands of files scattered across MacOS X UNIX-specific system directories not even visible in MacOS X Finder. However, thanks to the Open Source community, managing some UNIX software becomes much more easier and what is very important, safer.

Credits

All described here would not possible without great work of talented programmers - Christoph Pfisterer, Max Horn and all others, who developed port of the package management system from Debian/GNU Linux (one of the most powerful and advanced this days) for MacOS X. This port is called Fink. Also, I would like to thank Marc Liyanage who helped me with initial adoption of MySQL.

What is Fink?

Fink is a MacOS X port of package management system APT (Advanced Package Tool), package manager dpkg, and some front-ends for these tools from Debian/GNU Linux. However, this is only a part of Fink does for MacOS X. The greatest feature of Fink is that it creates its own UNIX directory tree (separated from MacOS X) where it keeps all its files (UNIX directory tree in brief described here).

Fink, as well as all UNIX software properly packaged with Fink (except Darwin/MacOS X version of XFree86 windowing system) do not install anything into MacOS X system directories (including "/usr/local") or alter any MacOS X system file(s). Thus, it is completely safe. If something goes wrong, one can just erase entire Fink directory without damaging MacOS X. However, please take into account that Fink do not tracks MacOS X native components, only software installed with Fink itself !

Installing Fink

If you have software from OpenOSX.com (Database, Web or GIMP CDs) or MacOSX.Forked.net you should remove them first! Some stuff from above mentioned companies is created with Fink and then repackaged for MacOS X.

Installing Fink is quite simple. Since Fink is a quickly evolving software, and you may use different version from main, installation procedure might be different, too. Fink distribution supplied with excellent manual, so just look into it. For version 0.3, launch Fink installer, wait until it finishes, then create text file named ".cshrc" (beginning with dot ".") containing line "source /sw/bin/init.csh" in your home directory Users -> Your Name (run "echo "source /sw/bin/init.csh" > .cshrc"), and finally run "rehash" in terminal. You will find "/sw" directory in the root level of your startup volume. This is where Fink keeps all its stuff.

Please note that Fink requires Apple Developer Tools and MacOS X SDK (both are parts of MacOS X Developer Tools CD available for free download) to be installed.

Upgrading Fink Core to the Next Release

Run "fink selfupdate" in terminal. This time (version 0.3 release) it is the only right way to upgrade Fink core in auto-pilot mode to the next release. You should not use MacOS X installer program and MacOS X Fink package for upgrade purposes (only for first time installation). To upgrade the rest of Fink installed components run "fink update-all". In order to upgrade to cutting edge release use technique described below.

Getting Started

  1. This article assumes that you have at least basic knowledge of UNIX and terminal commands.
  2. Almost all commands below must be executed in superuser mode (look here how to gain root access in MacOS X).

Please note that you need fast Internet connection to fetch source tarball(s) and Fink packages from the master site. Run "fink list" in order to see list of available Fink packages. To install (or upgrade) already precompiled Fink-enabled packages I may advice to use dselect (console based front-end to the Debian package manager dpkg). It is quite confusing for novices, but very capable and powerful. First of all, run Update command in dselect (or "apt-get update" in terminal) to fetch list of precompiled Debian-style packages which are usually referred as debs (but do not run Access, it may overwrite Fink settings!). Then, in Select section choose (with + key) whatever you what to install. And finally, run Install. The dselect will automatically resolve dependencies, download and install necessary components.

If you find dselect too cumbersome, just run "apt-get install package-name" in terminal (but do not forget to run "apt-get update" first !). If "apt-get install package-name" fails to install known package, it means that selected package is not available yet as precompiled deb from Fink master site, it have to be downloaded as source tarball and compiled locally. In this case you should run "fink install package-name". Please take into account that this process my be very time consuming.

Brief list of some Fink, apt-get and dpkg commands is available below.

Command Description Notes
fink list Lists available Fink packages At the time of Fink version 0.3 release 227 packages available
fink describe package-name Prints comprehensive description of specified package  
fink install package-name Downloads source tarball, compiles, and installs compiled deb package 1) Checks package dependencies; 2) might be very time consuming, try to use "apt-get install package-name" instead whenever possible
fink build package-name Downloads source tarball if it is not present and builds deb package No installation performed
fink rebuild package-name Works like "fink build", but replaces current deb package Automatically upgrades package with the newest build if another one was installed before
fink remove package-name Removes specified package Warning !!! Does not check dependencies (Fink version 0.3 and below), so be very careful, use "apt-get remove", "dpkg --remove" or "dpkg --purge" instead
fink selfupdate Updates Fink and its core components This is the only right way to upgrade Fink core to the new release!
apt-get update Fetches lists of available deb packages from master site  
apt-get install package-name Downloads and installs specified deb package Preferred method of installation, but do not forget to run "apt-get update" first!
apt-get remove package-name Removes package Preferred method of removal, checks package dependencies
dpkg --install package-name Installs selected deb package Checks package dependencies, but do not fetches dependent items
dpkg --remove package-name Removes package, but leaves configuration files Checks package dependencies
dpkg --purge package-name Removes package and its configuration files Checks package dependencies

man fink
man apt-get
man dpkg

Displays manual pages for Fink, apt-get and dpkg respectively  

What is the difference between "apt-get install package-name" and "dpkg --install package-name" one may ask? "apt-get install ..." will search for the specified package in the known package list and then will download and install it (example - "apt-get install mysql"), while with "dpkg --install ..." it is required to explicitly specify file name and path of the deb package (example - "dpkg --install /Users/Shared/mysql_3.23.43-2_darwin-powerpc.deb"). Both perform checking of package dependencies, but only "apt-get" will automatically fetch and install dependent items. In short, "apt-get" is preferred over "dpkg" whenever possible because it is higher-lever tool.

Another Fink Benefits

Fink may install some very useful Open Source UNIX software which is missing from MacOS X (or is just incomplete) like bzip2 (Burrows-Wheeler block sorting compressor with Huffman coding), wget (http/ftp recursive file retriever), tar (tape archiver which really works), Midnight Commander (very popular among Linux/FreeBSD users console based file manager), MySQL/PostgreSQL (powerful open source SQL database servers), and much much more, all this (except XFree86 installation) without altering MacOS X system directories.

Creating deb Packages for MacOS X/Fink

Full and comprehensive instruction available at Fink Web site, however, it assumes that you are an UNIX programmer. Below you will find some explanation how to quickly build your own deb without digging deep.

In short, you need to place source tarball and package description file into proper directories and run "fink build package-name". The top level build directory is "/sw/finks/dists". It contains 3 subdirectories (let's call them release status trees) - stable (final or stable releases), unstable (beta, testing or experimental releases), and local (releases built locally by administrator). Most likely all your home brewed software will go into local. Each release status tree contains two subdirectories - finkinfo (for package description files and patches) and binary-darwin-powerpc (for compiled deb packages). All source tarballs must be placed into "/sw/src/", or you may specify download URL in the package description file. Writing package description file from scratch may be rather tricky, so it is much easier to modify existing ones. For example, Fink 0.3 have been supplied with MySQL 3.23.42 without InnoDB, but I am needed MySQL 3.23.44 with InnoDB transaction safe table support. I have had to change just few lines in package description file (version #, revision #; and add compilation configuration parameter "--with-innodb"). You can download original and modified package description *.info files here.

Using Cutting Edge Version of Fink

Fink is being developed very actively, and the last packaged binary release may not be actually up to date. In this case you can install packaged release and then upgrade it to the cutting edge CVS version. In order to upgrade Fink core to CVS version run the following commands: "mkdir fink-cvs" (this will create temporary directory for Fink sources to be downloaded from CVS), "cd fink-cvs", "cvs -d:pserver:anonymous@cvs.sourceforge.net:/cvsroot/fink login" (you will be asked for password, just press Enter), "cvs -z3 -d:pserver:anonymous@cvs.sourceforge.net:/cvsroot/fink co fink" (it will take some time to download required files), "cd fink", "./inject.pl". You can remove temporary directory with "rm -R fink-cvs" or leave it for future updates from CVS. To update existing local Fink CVS sources run "cd fink-cvs/fink", "cvs -z3 update -d" and then "./inject.pl" to update Fink itself.

After you have upgraded Fink it is time to get catalog of packages available from CVS. Run "mkdir fink-cvs", "cvs -d:pserver:anonymous@cvs.sourceforge.net:/cvsroot/fink login" (you will be asked for password, just press Enter), "cvs -z3 -d:pserver:anonymous@cvs.sourceforge.net:/cvsroot/fink co packages", "cd packages", "./inject.pl". The only remaining step is to build and install packages - "fink update-all".

Q&A

Q. What to do if one want to rebuild an existing Fink package but with different compilation options?
A. Modify existing *.info file (increment revision number) and copy it into local tree, then run "fink build package-name", "dpkg --install package-name".

Q. How to tinker with stuff from the unstable tree (beta, testing, experimental, etc.)?
A. Copy desired package description (*.info) and patch (*.patch) files from unstable tree "/sw/fink/dists/unstable/..." to local one "/sw/fink/dists/local/..." and run "fink build package-name".

Q. The package I am looking for seem to be not available.
A. 1) Look in the unstable tree; 2) create it yourself.

Q. I am running "fink build package-name", but Fink tries to compile and build a lot of dependent packages, it will take hours or even days !!!
A. Fink build tool is not advanced enough yet and it does not check if precompiled dependent packages are available (at least this is true for Fink version 0.3). There are two solutions: 1) install existing package (if it exists, of course), with "apt-get update", "apt-get install package-name", then upgrade to your own version or revision; 2) manually install dependent packages with "apt-get update", "apt-get install dependent-package-name".

Q. The "apt-get install package-name" fails to install known package (error message - package not found). What to do?
A. Please read this article carefully again from the beginning. In short - it means that requested package is not yet available as precompiled deb. Use "fink install package-name" instead, it will download, compile and install package from sources.

Q. How to install XFree86 with Fink?
A. Excellent guide called "Running X11 on Darwin and Mac OS X" written by Christoph Pfisterer is available here.

Q. I have a problem with Fink and could not figure out how to solve it.
A. 1) Study Fink manual, other resources and be creative; 2) subscribe to the Fink mailing lists and ask there for help. However, please be patient and polite, and don't get mad if you won't get desired answer. Nobody is obliged to help you unless you paid for support contract. After all, you got Fink software for free.

Example - Building MySQL (with InnoDB) deb for MacOS X/Fink

  1. Download my package description files and copy them to "/sw/dists/local/main/finkinfo/". You cannot do it with MacOS X Finder, so run "cp mysql-3.23.44.info /sw/fink/dists/local/main/finkinfo" in terminal.
  2. Download MySQL 3.23.44 source tarball from www.mysql.com and copy it into "sw/src/" ("cp mysql-3.23.44.tar.gz /sw/src"), or specify download URL in package description file (I had tarballs hosted on my local workstation, so you may have to update download URL according to your setup). There is nothing to be changed in the source tarball itself, it may be used as is.
  3. Run "fink list". You should see "mysql 3.23.44-1 MySQL with InnoDB Support " in the list of available Fink packages. If you have another MySQL version already installed, Fink's list item will look like "(i) mysql 3.23.44-1 MySQL with InnoDB Support" ("(i)" stand as another version installed).
  4. Run "fink build mysql" and wait, it will take some time to compile MySQL and build deb package. MySQL deb will be located in "/sw/fink/dists/local/main/binary-darwin-powerpc/" directory, with symbolic link (analog of MacOS alias) in "/sw/fink/debs/" for easier navigation.
  5. Run "cd /sw/fink/debs/", "dpkg --install mysql_3.23.44-1_darwin-powerpc.deb". Please note that you do not need to run mysql_install_db script, it will be done for you at post-install stage.
  6. InnoDB support must be explicitly enabled in MySQL configuration file "my.cnf", for the smoke test just take main, copy it into "/sw/var/mysql" and set proper permissions ("cp my.cnf /sw/var/mysql/", "chown -R mysql /sw/var/mysql/my.cnf").
  7. In order to make MySQL data directory readable in the MacOS X Finder run "chmod go+rx+rx /sw/var/mysql".
  8. Launch MySQL daemon from terminal with "safe_mysqld --user=mysql &" (for 3.x version) or "mysqld_safe --user=mysql &" (for version 4.x) commands. Run "daemonic enable mysql" to create startup item for MySQL daemon.
  9. Take a look into "/sw/var/mysql/", if everything is OK you will find there InnoDB data and log files.
  10. You may build MySQL 4.x alpha version, too, but the package will be named mysql-alpha. You will need to rename source tarball and enclosing source directory accordingly.
MySQL 3.23.4x and 4.0 alpha do not shutdown cleanly due to incomplete POSIX thread implementation in Darwin/MacOS 10.1. This problem is fixed in MySQL 3.23.47 and later.

Downloads

MySQL deb for MacOS X is available for download here.


MacGuruHQ and MacGuruHQ Logo

Copyright© 1996 - 2002 by Andrei Verovski. All right reserved. Content may not be copied in whole or in part, through electronic or other means, without the author's permission. MacGuru Logo (blue penguin) by Sergey Orehov.