the Fink Wiki - User contributions [en]

Fink:Policy:Trees

2010-07-01T17:31:04Z

Dmrrsn: revise description of ssl linking

{{Fink_Header}}

== Linking Against OpenSSL ==

If the package currently has to be marked "License: Restrictive" because it [[Fink:Policy:Crypto|links against OpenSSL]], there is a workaround. The GPL specifies that it is OK to link against non-GPL software if it is a system-provided library. Since Apple includes OpenSSL as part of the base Mac OS X install, it can be linked against without license issues.

To do so, use the [[Fink:Policy:system-openssl|system-openssl-dev]] package.

== Tree Modularity ==

Any package in the stable tree must have all its (Build)Depends be satisfiable by packages in the stable tree. The idea is that a user could use only the stable tree and still still have a self-contained fink installation with no missing dependent packages.

Prior to July 1, 2010, any package that had a (Build)Depends that can only be satisfied by package(s) in section:crypto was required to itself be in section:crypto. This requirement was dropped when crypto was moved into the main tree.

You can pipe the output of 'fink list' into 'fink rebuild' to see if there are any packages with broken dependencies. The subsets are:
# stable/main
# unstable/main stable/main
(There used to be two more subsets, involving the crypto tree.)

Because some users might only have unstable/main, every package in the stable tree should also appear in the unstable tree, in either the same or a more recent version.

You can use fink's --trees flag instead of having to alter the <code>Trees</code> line in your <code>fink.conf</code> file for each subset of the distro, for example:

<pre>fink --trees=stable/main list -t | grep -v '\[' | cut -f2 > pkgs</pre>
edit <code>pkgs</code> to remove any known blockers:
* <code>libgems-rb*</code> and <code>rubygems-rb*<code> (-rb16 and -rb18 cannot be built concurrently)
<pre>cat pkgs | xargs fink --trees=stable/main,virtual rebuild</pre>

An alternative way to test for modularity and find all unsatisfied dependencies is:

<pre>fink --trees=stable,virtual list --format=dotty > foo.dot
perl -n -e 'if (/"(.*?)" -> "(.*?)"/) { push @{$need{$2}}, $1} elsif (/"(.*?)" \[/) {$have{$1}++}} END {foreach (sort keys %need) {print "missing $_ for ".join(" ",@{$need{$_}})."\n" unless $have{$_}}' foo.dot</pre>

{{Fink_Header}}

[[Category: Fink|Policy:Trees]]
[[Category: Fink_Documentation|Policy:Trees]]
[[Category: Fink_Policy|Trees]]

Server Administration

2008-09-26T19:07:01Z

Dmrrsn: howto restart the mail server

{{Fink_Header}}

== Backups ==

Backups are being done with Bacula. It deserves [[Admin:Backups|its own page]].

== DNS ==

We are running the default bind9 installation from debian. Domains are each in a separate file in <code>/etc/bind/</code> named after the top-level domain (finkmirrors.net, finkdeveloper.net, etc.) and are editable by anyone in the <code>bind</code> group (or root).

To add a finkmirrors.net record, do the following:

# edit /etc/bind/finkmirrors.net
# change the Serial to today's date
# add a distfiles.AIRPORT.STATE.CONTINENT record in the "distfiles (source tarballs) mirrors" section
# add a AIRPORT.STATE.CONTINENT record in the "bindist (.deb package) mirrors" section
# save the file
# restart bind: <code>/etc/init.d/bind9 restart</code>

== Mail ==

We run a mail server. It really really needs to be documented. (In particular, we are running squirrelmail, websieve, and god knows what else.)

In particular, the following URL's are useful:

<https://www.finkdeveloper.net/squirrelmail/src/login.php>

<https://www.finkdeveloper.net/cgi-bin/websieve.pl>

Authorized persons can restart the mail server with <code> sudo /etc/init.d/cyrus2.2 restart</code>

== Creating New Users ==

# add the user to LDAP
#* details go here
# create their SVN experimental directory
#* <code>svnadmin create --fs-type fsfs /var/lib/svn/users/'''username'''</code>
#* <code>chown -R www-data:www-data /var/lib/svn/users/'''username'''</code>
# create their .users.finkproject.org web site
#* add to /etc/bind/finkproject.org
#** <code>2006XXXXXX ; serial</code>
#** <code>'''username'''.users IN CNAME finch</code>
#** <code>/etc/init.d/bind restart</code>
#* add to apache2 config
#** <code>sed -e 's,@USER@,'''username''',g < /etc/apache2/users/user.template > /etc/apache2/users/'''username'''.user</code>
#** <code>/etc/init.d/apache2 restart</code>

{{Fink_Header}}

[[Category: Fink|FinkProject]]
[[Category: Fink Structure|FinkProject]]

Fink:BinDists

2008-07-10T12:59:02Z

Dmrrsn: more about VERSION

{{Fink_Header}}

Periodically, Fink releases a "binary distribution" or "point distribution", traditionally numbered according to the Darwin version number. This consists of several things:
*a source release from which fink can be bootstrapped, which includes all of the packages in the distribution (such as fink-full-10.5-0.9.0.tar.gz)
*a release of a tarball of all the .info files from the distribution (such as dists-10.5-0.9.0.tar.gz)
*periodic subsequent releases of tarballs of .info files from updates to the distribution (such as dists-10.5-0.9.0.1.tar.gz)
*binary .deb files for pre-compiled versions of all of the packages in the distribution, for all currently-supported architectures
*periodic updates to the set of binary .deb files, as needed (in conjunction with updates such as dists-10.5-0.9.0.1.tar.gz)
*binary installers for the distribution (such as Fink-0.9.0-Intel-Installer.dmg and Fink-0.9.0-PowerPC-Installer.dmg

Fink makes pre-compiled versions of its "stable" packages available for automatic installation (when the package license permits). Many of these packages were released under the Gnu Public License (GPL), and the Fink project takes its obligations under the GPL quite seriously. See [[Fink:ArchiveBrowser|Archive Browser]] for more information.

Here are a few tips for the Fink Developer team:
* the set of .info files in a distribution should be tagged with a cvs tag such as release_0_9_0
* the set of .info files in the updates to a distribution should be tagged with a cvs tag such as release_0_9_0_1; these updates should be cumulative, so that release_0_9_0_2 contains everything from release_0_9_0_1 plus additional updates. (The pdb must be configured to grab both of these cvs tags and get fink to merge their contents when indexing them.)
* note that any of these tagged distributions must include the files inject.pl ,DISTRIBUTION, VERSION, README, and stamp-cvs-live, and must also include at least one .info file (in the appropriate heierarchy). The tarball-rolling script scripts/srcdist/dist-module.pl will replace stamp-cvs-live with stamp-rel-$version.
* the VERSION file in an update release such as release_0_9_0_1 should still contain the string 0.9.0-cvs, not reflecting the update. The tarball-rolling script scripts/srcdist/dist-module.pl will insert the correct value for users obtaining (and injecting) the tarball.

{{Fink_Header}}

[[Category: Fink|BInary Distributions]]
[[Category: Fink Structure|Binary Distributions]]

Fink:Roadmap:0.29 new features

2008-06-24T00:11:34Z

Dmrrsn: new page

{{Fink_Header}}

===New features in fink 0.29===

* separate Restrictive packages

===Bug fixes in fink 0.29===

*

{{Fink_Header}}

[[Category: Fink|Roadmap 0.29 New Features]]
[[Category: Fink_Roadmap|0.29 New Features]]

Fink:Major New Feature Plans:Modernizing the Upgrade Process

2007-10-24T13:42:12Z

Dmrrsn: wikify

{{Fink_Header}}

Starting with fink-0.27.7, fink has a new method for handling upgrades which change the distribution.

First, each fink.info file is being given a Distribution tag, and its own Revision increment. This way, when the underlying OS is upgrading, fink will recognize that it needs an update.

Second, to handle the problem that older finks don't recognize the Distribution tag, each fink.info for distributions 10.5 and later is being wrapped in Info4. Old finks which do not recognize the Distribution tag also can't handle Info4, and will refuse to process those files. Thus, fink-10.4.info with no Info4 and fink-10.5.info with an Info4 can coexist in the same finkinfo tree, and an older fink will only see the former. Once it has updated that one, it will see the latter too and -- if on 10.5 -- will update again.

Third, fink's postinstall.pl now compares the Distribution value stored in fink.conf with the one obtained by querying the system, so that it knows if we are upgrading. If we are upgrading, it stores the old Distribution value in a new fink.conf entry "UpdateFrom", for later handling by SelfUpdate.pm. It also may, depending on which upgrade is being performed, store a list of packages in "UpdatePackages". (These are packages, such as perl586-core during a 10.4->10.5 upgrade, which need to be installed immediately to prevent missing dependencies.) The info and patch files for any update-packages are being supplied with fink itself, and at the end of postinstall.pl they are copied to %p/fink/dists/local/injected (where they will subsequnetly found when SelfUpdate.pm needs them.)

In a related change, fink's postinstall.pl now processes the SelfUpdateTrees information, modifying it appropriately if we are upgrading, and storing in fink.conf. (If SelfUpdateTrees is not present but the old variable SelfUpdateCVSTrees _is_ present, it uses the latter to obtain information, but stores the information in the new variable and removes the old variable.)

The presence of an "UpdateFrom" variable indicates to fink selfupdate-finish that there is work to be done. However, this has not yet been implemented.

The one part of the work in SelfUpdate.pm which _has_ been implemented is installing the "UpdatePackages" -- these are installed along with "essential" and "important" packages during fink selfupdate-finish.

{{Fink_Header}}

[[Category: Fink|New Feature Plans Modernizing the Upgrade Process]]
[[Category: Fink_Roadmap|New Feature Plans Modernizing the Upgrade Process]]

Fink:PackageDatabase

2007-04-11T17:40:23Z

Dmrrsn: PDB has moved

{{Fink_Header}}

One of the most heavily-accesssed portions of the fink website is the searchable package database. This portion of the website has its own URL pdb.finkproject.org for historical reasons (it was formerly hosted by opendarwin.org), but it is now running as part of the main website. A mysql database contains the data, which is updated at two times: first, when a new binary distribution is released, the data from the distribution is inserted into the database, and second, several times a day an automatic update is made from the fink CVS repository and the pdb scripts convert this into mysql data which is then put into the database (replacing previous information in the database).

An effort to enhance the package database is documented at [[Fink:PackageDatabase:NewDesign]].

{{Fink_Header}}

[[Category: Fink|Package Database]]
[[Category: Fink Structure|Package Database]]

Fink:SourceForge

2007-03-10T22:36:10Z

Dmrrsn: website not here

{{Fink_Header}}

Fink was brought to sourceforge.net a few months after its founding. We've valued our relationship with this service over the years (and vice versa: we were one of the earliest "Project of the Month" at sourceforge), but as fink has grown, parts of our operation have migrated away from sourceforge. Currently, sourceforge provides us with the following services:
* Accounts for Fink developers, and for the fink project itself
* Hosting of mailing lists
* CVS (including full CVS tarballs compiled daily; including read-only anonymous CVS access for our users; and including a nice web interface to CVS)
* The sourceforge release system, used to release the files of fink projects and to host backups of source files from other opensource projects
* Bug tracker (and related trackers: for submitted packages, for patches, and so on)

{{Fink_Header}}

[[Category: Fink|SourceForge]]
[[Category: Fink Structure|SourceForge]]