Changes

resurrect the debian branch

make mysql-server recommended

add miguel, update years

don't install apache.conf, install docs and examples properly, install files properly

files moved

rewrite installer

 1. don't install drushrc
 2. don't install apache config (hostmaster-install does that)
 3. install directly in /var/aegir again, as aegir assumes that

update readme file

update TODO

fix aegir location

update changelog with new release and latest change

add a few debconf questions

fix makefile whitespace and version numbers

head towards a release

fix docs and examples

update with the new violations of FHS

add missing debhelper depends

forgot to source debconf

fix ucf call sequence

add debugging

fix templates and config

handle existing manual provision install gracefully

add minimal rules file

default the site name to the FQDN

prompt for the db user and password, pass that insecurely to the backend

don't bother the user with prompts, since we already do this in debconf

install the right version of the frontend

have the makefile path be relative to the install script

fix typo

fix permissions on sudoers file to please sudo

configure apache properly after installation, fix permissions

update the todo

try to set the password as seen if we successfully installed

fix case where aegir config already exists

try to support upgrades

mark the mysql pass as seen in any case of success

disable debugging, this works well now

we're checking for existing installs and are not prompting for
passwords now

improve changelog

fix old platform parsing regex

explain how to migrate to this package

don't ask questions if there's already an install

don't ask questions on upgrade

upgrades work!

more documentation on the next steps

add a command to parse --backend output

i got tired of stabbing my eyes with inline JSON output so I wrote a
very very simple command that just parses STDIN as a --backend output
command. this can be useful for debugging when calling backend commands
without thinking of removing --backend, or in my case, to pass data
through STDIN as JSON, but still have a proper output.

pass the mysql root password security through STDIN

always prompt for the password on reinstalls, give feedback to user

make the password priority high so it's asked in most cases

forget the value of the db password in config

we need to do this *before* we prompt (db_input)

clarify information given to the user on existing installs

describe required work for safe_shell_exec()

initialise $pipes array for clarity

don't chown non-existent dirs

do not Depend on later sudo

we now inform the user if sudo cannot be configured automatically,
instead of just failing to install because of silly dependencies.

update changelog for beta2

specify source format

fix typo

fix version parsing

add info file

fix version parsing

check if the platform is really the same, don't just take the existence of the directory for sufficient

update install.sh.txt to fetch the right version of provision

switch to fetching hostmaster from tag rather than branch

upgrade wording, vcs repos

update todo

update watchfile to point to drupal.org

remove unused config file

remove unpackaged docs dir

announce 1.0~rc2

remove frontend installer

configure git-buildpackage

update standards version, no change

do not leak the credentials in the environment after we have used them

update changelog

readd frontend installer as a separate package

fix version number in .info file

call su with an explicit shell as we may not have one

#730424 - only use apache_setenv if it exists, fixes nginx support

use drush --pipe argument to make output easily parsable

compare canonical paths when checking for upgrades

make an 'aegir' metapackage, clarify the other packages descriptions

announce 1.0~rc3-1

fix descriptions to please lintian

update the .info version

allow easier debugging from outside this script by respecting DEBUG=yes

release 1.0~rc3-2 with the DEBUG command

don't fiddle around replacing the content of the sudoers file, we just
support apache2 now

abort earlier if provision is already installed

depend explicitely on apache2, apache has been removed from debian
ages ago

announce -3

make the sample sudoers useable out of the box

fix typo

pass the makefiles on upgrades also

fix version number *again*

announce -4

cleanup empty platforms on upgrade

announce -5

add more debugging

more debugging

announce -6

prompt for makefile anyways

-7

fix long lines

fix site prompt and some priorities

make sure we prompt for the makefile if it was never seen

anounce 1.0~rc3-8, hopefully the last

fix typo in sudoers sample file path

silence warning on new installs

announce 1.0~rc3-9

move apache config unregistration to remove, as per webapps policy

simplify sudo config file handling in postrm

detail policy compliance issue, add signature security issue

add changelogs for -9

add config directory since we now use it in postinst

announce 1.0~rc3-10

change version information for release 6.x-1.0-rc4

make a test release on the 1.x branch

fix platform cleanup after upgrade

provision-delete the platform instead of manually

rm -rf on the platform is a very bad idea because it breaks the apache
config and then it fails to restart. provision-delete cleans up
properly.

replace the test release with the rc4 release notes

autodetect drush, to be compatible with system-wide drush installs

fix command to find platform alias

roll out 1.0~rc4-2

remove hosting-pause call, we'll put it back once everybody hit 1.0 at least

call hostmaster-resume again

hosting-resume doesn't exist in pre 1.0-rc4, so this fails during upgrades

we'll remove this once everybody hit at least 1.0-rc4 (or say 1.0)

announce -3

fetch from hostmaster 1.x

really ask the darn makefile question on upgrades from manual

fix message and make this work every time

announce 1.0~rc5-1

require at least drush 4.3

Revert "change version information for release 6.x-1.0-rc5"

This reverts commit 321b1522920661baa82ced34adb1ea291deb86e6.

Client directory with symlinks should have group read access granted - by grace

default release on the branch is 1.x, so that's what we upgrade to

change version information for release 6.x-1.0-rc5

Revert "change version information for release 6.x-1.0-rc5"

This reverts commit b269e71aa935eaf589af3f0c47ed9d951da2e656.

1119850 fix Drupal 7 support by sending the right server variable in db cloaking

1083710 by tbosviel - fix php 5.3 warnings in DNS

remove the AEGIR_HOME variable that was adding confusion and bugs

we have no HEAD anymore, but 6.x-1.x

document _provision_file_check_location()

port our recursive_delete() function to drupal 7's

improvements;

 * we close directories as we go
 * coding style

1119968 do not follow symlinks in recursive delete

that way we do not need to check if the path is in the right directory
at all, we just remove the link, who cares if it points to a place
outside the site dir if we do not dereference it?

properly carry errors through multiple directory entries in recursive delete

comment formatting

Issue #1118286: specific config functions for each DNS service.

try to upgrade drush even if it's the path

we rely on readlink -f, which is not really cross-platform, but the
alternative is not supporting symlinks properly...

change version information for release 6.x-1.0-rc6

announce rc6, no change in debian package

fix the download type in the makefile

announce rc6-2

rc7-1

fix permissions on config dirs

add bugfix #1123276

depend on git

document how we can also install the frontend automatically now

got approval from daniel to quote him

typo

set the varialbes back to how they were. should be a real revert here but i'm in a rush

#1123176 by sanduhrs - switch to mktemp instead of tempfile in upgrade script to retain CentOS out-of-the-box compatibility

Issue #1027358 by Steven Jones

Change the method of detecting if we should add the Aegir API version to the settings.php file.

993944 rollback http configs in site installs failures

#1016890 by joestewart - drop setContext function in the cluster service to fix the cluster feature

rename get_certificate_ip() to assign_certificate_ip() before I go insane

this is to distinguish with the get_ip_certificate() (!!) function
which is a real "get" whereas assign *does* something

1081536 rollback platform apache configs

when drush make fails, the apache config is already created and wasn't
cleaned up

we hook into the rollback by adding a new magic hook through
command_invoke() that will cascade through the current context and call
the right rollback hook.

we implement one for the platform, but we probably need to implement one
for servers and sites too.

Revert "1081536 rollback platform apache configs"

This reverts commit f7a071135d626bcef4879dc7e8687fd888d8383b.

The problem with this approach is that if we have sites in the platform
and fail to verify it, removing the apache config would actually break
the platform even more. We do not want that.

What we want is a provision-install on platforms that can be rolled back
properly too. It's harder than this.

try to rephrase things to provoke jenkins

more docs in the sudoers file

deprecate most of this file

remove a reference to UPGRADE.txt which doesn't exist anymore

Issue #1074966: fix warning in command line mode.

White space changes in bind_service.inc.

Fix for Nginx configuration: if the file does not exist, send 404 error instead of back to parent location.

Add support for AdvAgg module in the Nginx configuration.

Add support for robotstxt module in the Nginx configuration.

documentation fix in release.sh

change version information for release 6.x-1.0

announce 1.0-1

upload to stable

put back in unstable, just doing some tests here

activate the required modules SSL and rewrite

make all maintainer scripts respect DEBUG

add missing mod_php dependency (!!)

announce 1.1-1

update dependencies to include unzip, php5 and recommend rsync and gd

thanks to Jamie McClelland for spotting the missing unzip dependency

use more standard DPKG_DEBUG flag for debugging

#1149172 - do not destroy old platform on upgrades

announce 1.1-2

allow also the use of DEBUG to enable debugging

this will be reverted on the 2.x branch, as it is deprecated. this was
mistakenly removed from 1.x so it will be merged there.

keep supporting DEBUG

keep supporting the DEBUG flag, it is now deprecated, and DPKG_DEBUG is
still supported.

announce 1.1-4

support subdirectories with spaces in them

fix debian build

remove phony targets and add jenkins build code

rename jenkins targets

fix variable in autobuilder

fix makefile shell call (duh)

try to fix funky makefile escapes

try to move build within the jenkins workspace

rebase to debian

build the debian packages as a binary, not source

fix auto build

try to build auto on a temp branch

try to simplify by using dh7 rules

update standards version, changelog

release 1.1-5 for real and tag

trash the tmpdebian branch at the beginning of the build

fix auto build version number to reflect upstream changes

otherwise this always looked like 1.1

simplify autobuild target by factoring in more assumptions

fix version detection, add show-version target for testing

do not clear the db_password prompt in config

That's beacuse it may be preseeded, and anyways it's already cleared
in postinst

announce 1.1-6

upload official packages directly to testing

Try not doing funny things to the db_password.

Revert "Try not doing funny things to the db_password."

This reverts commit 9396b6b2222f2b9f3a2708d1dc706f2891594bee.

Try not resetting the db_password.

Add a touch of debugging.

Remove the nasty debugging statement.

autobuilder fixes from drush:

 * upload to unstable
 * fix commitlog to indicate the jenkins build tag
 * make PGP key customizable

update debian/changelog

announce 1.2-1

make sure the email/sitename is prompted

this is a test commit, but apparently the way things are right now,
those things are never prompted even though aegir/site is high
priority.

try to fix mysql pass double-prompt

somehow we get prompted for the mysql password twice

make email medium priority so it's asked

the rationale here is that it's then easier to install aegir for a client or somebody else than you

announce 1.3-1

fix download url in copyright file

announce 1.4-1

fix git-buildpackage configuration to follow new versions

check if aegir is bootstrapping properly after upgrade or install

this will make sure things fail properly, see #1300216

Try to install Aegir from the users home directory.

Refs #1309080.

Update the directories we need to make.

Try to copy the tests in a better way.

Try to fix the copying not working.

Remove the bash conditional, and use the test command. Magic.

Put the bash conditional on a single line.

require the drush_make package

this also disables (chmod 0) the manual drush_make install so that
drush doesn't freak out on the duplicate code copy.

it's drush-make, not drush_make

match versions in dependencies

this will allows us to just do apt-get install aegir to upgrade all three packages

move manual provision detection code to preinst to limit the damages in case it happens

do not chmod0 drush_make, just abort if it's found

fix syntax error that oddly enough doesn't crash the install (?WTF?)

duh - forgot to define VARLIB=/var/aegir

1.5-1 announce

fix distribution

1328316 hotfix: do not trash the aegir configs on failed upgrades

announce 1.5-2

do not build from the upstream branch, but the tag

Updating Debian changelog for release 1.6-1

first implementation of the frontend slave debian package

add missing platforms dir, warn the user of necessary reboot

fix examples path

use VARLIB properly

add postrm

do not edit config files if no value is provided (default)

inform the user we do not edit if no value

remove debuggin output

move umount to prerm to remove warning from dpkg

fix missing bits in postrm

fix version number so it's lower than autobuilds

add missing mysql-client dependency to slaves

make this a native package

while Aegir is desgined to run on other systems than Debian, it is
currently *not* uploaded into Debian itself, and therefore there is
only one source of packages, debian.aegirproject.org.

this will mean we will need to make a new release every time we want to publish changes to the Debian package, but this annoyance outweights the trouble of having to maintain 4 different branches (1.x, 2.x and debian, debian2x) to have a 2.x debian package.

once we start looking again at sending the Debian packages directly into the main Debian archive, we can re-create the debian branch and make non-native packages there. for example, 2.x could eventually have no debian/ directory and the debian branch would be solely for the 2.x release...

in the meantime the debian branch is not to be used.

remove watchfile

now that we are a native package it doesn't make sense

mark the debian branch merge

rebase instead of just pull

edit the changelog in the release.sh script

fix my email address

we are native, no sense in using quilt anymore