Fix un-initialized member , reported by Coverity and also at http://root.cern.ch/phpBB3/viewtopic.php?f=15&t=18309

gProof is not a member of TProof !!!!

Note: R__ASSERT are never elided and can be use for check that should
be done in both debug and optimized builds.

Installation of prepareHistFactory. Fixed the logic in selecting what needs to be executed as test having precedence the target names. Avoid  include ${ROOT_USE_FILE} in tests.

Added installation of headers of histfactory and tmva to 'headers' component. This fixes the problem seem by Lorenzo that some headers were not copied to the binary directory.

The changes are all related to the use of gendered or binary language in
the comments, messages, and documents (e.g. referring to the user explicitly
as "he" or "he/she").  I searched for thse instances and changed them to
the gender-neutral singular they (http://en.wikipedia.org/wiki/Singular_they).
In the process I fixed quite a few typos that would show up in the
auto-generated documentation.

By Jean-François Caron jfcaron@phas.ubc.ca).

Fix also setting fill/line colour for the contour

see http://root.cern.ch/phpBB3/viewtopic.php?f=15&t=18038

rootcling will write out ROOT files; that does not work for libCore and RIO
which instead use rootcling_tmp.
All other libraries use rootcling -rootbuild.
Rename ROOTCLINGTMP to ROOTCLINGSTAGE1; rootcling -rootbuild is ROOTCLINGSTAGE2.

fix commit 127aa29a which prevents delete under certain circumstances

The construction of a RooKeysPdf is now much faster for large data sets.
This is achieved by carefully analysing where the different Gaussians
that the code superimposes can actually contribute. If a contribution
is smaller than the machine precision, the code stops trying to put in
those contributions (it would be lost in rounding anyway), which leads
to roughly 40 % faster construction of the class for large data sets.

Fix calculation of extended term for weight2 likelihood. This fixes the problem observed with getting the correct errors in weighted extended likelihood fit.
See http://root.cern.ch/phpBB3/viewtopic.php?f=15&t=16917&p=75834#p75834

Note : the version of RooNLLVar on the master was different than 5.34 patches

- Fix the HypoTestInverter::Rebuild to re-set the parameter to the same initial value before generating the toys. Apply also several debug improvements

- Apply some fixes for ProofLite
- have now a deterministic seed making the result reproducible
- allow for setting correctly the number of workers when running in ProofLite

Improve StandardHypoTestINvMacro for the rebuild.
Add an option to specify the configuration used for rebuilding the limit distribution

Use modified formula for tq-tilde only when qmu is significant larger than qmu_asimov. Avoid in particular the case when both are very close to zero.

- Fix in case of Poisson constraints or Poisson models to set the option noRounding to get correct real value for the Asimov data sets
- Remove unneeded TCanvas settings in histfactory

In this way the model can be written correctly in an XML file.
Do come clean-up of code in PrintXML methods
Update documentation of Channel.h and Data.h update meaning of histoname and histopath

This fixes ROOT-5480

This fixes ROOT-5481

Print only in debug mode

Fix a crash in analyticalIntegral when rangeName is null. This fixes the segfault reported in ROOT-6031

Fix a problem in PrintXML when not passing a directory (default case)
Print also  histogram address in Data::Print()

Remove usage of TRef in HistFactory. It is not needed. Replace with a simple wrapped class to manage histogram pointer
(RooStats::HistFactory::HistRef)
This should fix a bug reported by Swagato
Fix also memory leaks created by cloning histograms in histooworkspace
Add a histfactory test from Christian (stresshistFactory ) in $ROOTSYS/test