User has to choose a custom installation.
There are two choices:
- Start at boot as "daemon" (enabled by default)
- Start at boot as "jenkins" (disabled by default)

If user disables the first and enable the latter, the installer
will create a new user "jenkins" and new group "jenkins"
(if they do not exist), chown JENKINS_HOME to jenkins:jenkins
and set up launchd job to run Jenkins as "jenkins"

Also fixed some permission errors that PackageMaker complains about
when building the package.

Also PackageMaker xml file reformatted to be more readable

Fixed a bug in Winstone where a failure during request processing can
put Winstone out of sync with the next request.

if the check fails miserably, report the error accordingly.

Fixed hash DoS problem.
(cherry picked from commit 5ef902bb)

Conflicts:

	changelog.html

Fixed hash DoS problem.

The closures sent from the CLI client should carry over its
authentication. Test is written in the SSH cli auth module.

Add some indirection to allow the promoted-builds plugin's
RedeployBatchTaskPublisher to override getMavenBuild() and return the
promotion target's MavenModuleSetBuild object.

Also requires a fix in promoted-builds.

Signed-off-by: Michael Smith <msmith@cbnco.com>

- I feel uncomfortable with a string field used as an enum.
- I feel uncomfortable with parameters that have some restrictions about
  the way they can be combined.

My change is bit more verbose but it uses less Jelly and I think it's
more in line with the architecture.

1) Installer checks Java is installed before proceeding.
   It also tells how to get Java installed.

2) Installer opens http://localhost:8080 when installation is finished.

[FIXED JENKINS-12119] rolling back a portion of 46998ba0446834de0c725c32b328ace7ce844f6a as this is binary incompatible change

The ZipArchiver was visit()'d with a relative path that uses the system's path
separator, but zip archives require a forward slash for platform compatibility.
This fix replaces the platform-dependent File.separatorChar with a forward
slash.

user name

Don't let the partially loaded Jenkins cause a possible data loss.

At some point after this code was originally written, I/O in remoting
became asynchronous. So we need to sync with that before we swap out
the new log output stream, or else you cut off stuff at the wrong
moment.

Normally the command transfer in remoting is slow enough that
you get more or less synchronized output anyway, which is why
this poblem remained unnoticed.

But if the uncliamed buffer gets large, write operation will take a long
time, which makes this issue more likely to show up.

This fix address this problem by doing a proper sync.
The recycling of ByteArrayOutputStream is also disabled because it makes
synchronization tricky.