Downgrade some of the logging messages in the ConfigService - they weren't necessarily errors and could be confusing. Re #1161.

Fix various memory leaks (mostly only in tests) and other problems flagged up by valgrind. Re #1154.

Modify test run scripts to properly take into account test failures that do not retrieve std output. Re #100

In Interpolation when outside table interpolate linearly instead of then taking endpoints.  
Refs #967.

Seems like this has been there from the very beginning of one of our first algorithms! Amazing that no-one's spotted it before now! It's in other places too (fixes to follow). Re #1159.

Improve the speed of LoadRaw in the presence of multiple periods, particularly in the case where the monitors are loaded separately where memory usage is also significantly reduced if there are large log files. Also fixed a minor bug in the attachment of period logs to workspaces. Re #1154.

<parameter name="somefunction:Alpha" type="fitting"> 
  <value val="250.0" /> 
  <min val="80%"/> 
  <max val="120%"/>
</parameter> 

results in Alpha being constrained to sit between 250*0.8=200 and 250*1.20=300.

Wiki http://www.mantidproject.org/InstrumentDefinitionFile updated.

Refs #967.

Update RPM build scripts to also install sip and PyQt4 into the Mantid bin directory so that they are not required on a user's system. Re #100

Code in preparation for allowing percentage constraints and further unit testing of formulas.  Refs #967.

MantidPlot scripting autocomplete now inspects objects to update auto complete information after each execution. Required that each tab + the interpreter have different dictionaries so that the autocomplete only sees information 
for the current script. Re #713

Updated build scripts to give better notification if a test dies without actually running anything. This was happening with a test that threw in a constructor and cxxtest doesn't guard against this.
After some digging it seems cxxtest::TestSuite has two functions: setUp and tearDown that can be overridden to setup and cleanup before and after a test respectively. From now on we should use these 
functions instead of relying on Constructors/Destructors to initialise and cleanup our test classes. Re #100

Don't bomb out when geometry fails to find track segment, just skip the element instead. Error induced by this expected to be negligible. Re #1123.

Correcting some error estimates: preventing negative errors and stopping errors from being set to zero when the Y-value is zero. refs #841

Added RTLD_GLOBAL flag to dlopen on POSIX so that when a library is loaded then all of the symbols defined there are available to the calling application. I was seeing problems with some dynamic casts relating to this. Re #1157

trying to be constructed. Re #1156

Add an algorithm to calculate absorption corrections for a generic sample shape. Algorithm name is AbsorptionCorrection, the class name is AnyShapeAbsorption. This also required changes to the Sample and Object classes because the sample shape object wasn't being copied properly. Re #1123.

<parameter name="somefunction:formula" type="fitting"> 
  <formula eq="100.0+10*value+value^2" unit="TOF"/> 
</parameter>  

Refs #967.

A bug fix in GroupDetectors2 and making the Excitations interface create and run a single Python function refs #1137 closes #1148