This prevents a NullPointerException from being thrown, and allows the
socketNodeClosing method to be called for the SocketNode.

The constructor creates an ObjectInputStream, and this blocks until the stream
header is received from the client.

If the client doesn't send a stream header, construction of the SocketNode will
hang, preventing SimpleSocketServer from accepting new clients.

This change moves the construction of the ObjectInputStream into
SocketNode.run. As SimpleSocketServer runs this method in a new thread, a rogue
client won't prevent subsequent ones being able to connect.

amendment for LOGBACK-644: suppress errors for nonexistent resource includes

Support for elapsed time to jetty and tomcat access logs

* changed direct access of timestamp field to use getTimstamp() instead inside of calculateElapsedTime.

Truncates multi-line subjects to the first line to prevent an SMTP error.

fixes #147

detachLoggerContext removes LoggerContext from ContextSelector. BUT
context.getLogger will create it again. So it's much more proper to
log at first and detach context after that.

fixes #51

Fix for LOGBACK-754

A NullPointerException occurred when setting a JUL logger's level to
null, which should have reset the logger level to the parent's effective
level. The null level was incorrectly propagated to context listeners,
leading to the NPE.

This patch propagates the logger's parent effective level.

Support suppressed exceptions

Check already exists inside called method

* Added support to the tomcat Server adaptor for correctly getting request start time. (Includes additional dependency on tomcat-coyote)
* Update layouts.html to document the addition of elapsedTime/D to the pattern support for IAccessEvent
* Modified AccessEvent.java to calculate the elapsedTime at construction of the event, instead of repeatedly in the getter.  This allows the value to be more easily seen in a debugger, and cleans up the code slightly.

LOGBACK-832 - set parent classloader to avoid janino classes not being found

Use new instance of RecoveryCoordinator in each test. Otherwise, the timestamp
of the next call to isTooSoon() could happen well after the backoff threshold
on slow machines (such as small nodes on Cloudbees Jenkins), causing false
test failures.

This also renames and refines the individual tests.

The slack factor for non-Sun JDK performance test in PackagingDataCalculatorTest
was not lenient enough for the small virtual nodes on Cloudbees Jenkins.
Some people were also reporting test failures here by a few percentage points
(on their slow development machines).

Bump slack factor from 10 to 15

fix for LOGBACK-842

Modified the regex patterns in RootCauseFirstThrowableProxyTest and
ExtendedThrowableProxyTest to find any packaging info in the throwable's
stack trace, including inexact matches.

On Cloudbees Jenkins, these tests were failing because they were looking
for an exact match of the JUnit package but only found an inexact one.

debug improvements to Gaffer

(cherry picked from commit bd88d3ed580a4f60dc54c3293eb42aee5ea73148)

Fix exception for logger name with '.' and '$'

LOGBACK-384 - Mixing periods and dollar signs in a logger
name causes IllegalArgumentException from LoggerFactory.getLogger

Fixing the logic to treat '$' and '.' at same level as per
comment [1]. Also adding a testcase

[1] http://jira.qos.ch/browse/LOGBACK-384?focusedCommentId=13105&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-13105

Use append instead of concatenate

Avoid NPE during JNDI lookup

Fixed an issue where an extra line of the stack trace was being printed by the ThrowableProxyConverter when specifying a length option.

The DataSourceConnectionSource.start() method opens a new connection
only to determine whether or not the discoverConnectionProperties()
should be called. This is unnecessary since that method will perform
the same check and issue a very similar warning. A more serious
problem is that the connection is never closed.

This commit removes the use of this extra connection completely. I
think this is the last issue in need of a fix before LOGBACK-376 can
be closed.

Use append instead of concatenate

Avoid NPE during JNDI lookup

LBCORE-179: Fix regarding incorrect stack trace depth when specifying %throwable with argument

Fixed JNDI lookup of non-String entries