...and add a test.

The new argument is to make TTreeProcessorMT consistent with other
multi-threading interfaces, namely TThreadExecutor. This is the last
ingredient required to fix ROOT-10561.

The new release includes some improvements in Reverse mode:
  * Reduce the quadratic cloning complexity to linear.
  * Support variable reassignments pontentially depending on control flow.
  * Support operators `+=`, `-=`, `*=`, `/=`, `,`, `++`, `--`.
  * Allow assignments to array subscripts.
  * Support nested assignments in expressions `a = b * ((c ? d : e) = f = g);`
  * Enable differentiation of for-loops

See more at: https://github.com/vgvassilev/clad/blob/v0.6/docs/ReleaseNotes.md

This patch enables us to upgrade to llvm9. Clad supports from clang5 to clang9.

Don't generate dictionary to throw it away when called with -r. This
patch should speed up rootcling -r and simplify some of the conditional
logic.

Quotes are necessary in case the Python2_VERSION variable is not defined,
to prevent a missing operand in the expression that compares the
version.

The exclude clause was a workaround travis-ci/travis-ci#4681 but it causes
travis-ci/travis-ci#8536 in recent travis.

...to make it clear that we only use these functions in this same
source file.

Fix for https://sft.its.cern.ch/jira/browse/ROOT-10654 and https://sft.its.cern.ch/jira/browse/ROOT-10599

Co-Authored-By: Stephan Hageboeck <stephan.hageboeck@cern.ch>

Internal logic has been adapted to support the case in which TTreeProcessorMT
is acting on a TChain and the trees in the TChain do not all have the same name.
Importantly, the fTreeName data-member changed name to fTreeNames and is now a
vector of strings rather than a single string.

This fixes ROOT-10178.

Previously it would retun the parent directory in case where a sub-directory creation was requested.

See https://root-forum.cern.ch/t/subdirectories/15665/11.

The reason is explained in https://sft.its.cern.ch/jira/browse/ROOT-10643

Old CINT was not able to parse them correctly

Old CINT was not able to parse includes, therefore protection was required

CINT was not able to parse them correctly

These includes may be used in dictionaries and therefore should
be available already in core libs

In TEntryList::Remove and TEntryList::GetEntryAndTree the error case of GetEntry (received
indirectly by TEntryList::Remove) was not treated and consequently 'undefined' behavior
followed.  stressEntryList triggered a crash in GetEntryAndTree when reading random
memory to fill the input TTree led to a situation where a TEntryList was created on a
large chain but then applied to only a subset of that chain __and__ the selection/random-number
was such that none of the entries of the subset was selected (but some of the other part of
the large chain was selected).  This lead GetEntryAndTree to call GetEntry in a state that
it returned an error code (-1) and left the TEntryList in a state unexpected by GetEntryAndTree
(namely that fCurrent is nullptr even though fLists is not nullptr)

Due to the previous random memory address read, we ended up testing a new content layout that
we are reproducing now with a 'fixed' cut.  In this test, we first create a TEntryList
from a large chain but apply it to a small chain but none of the entries in the small tree was selected
then it triggered a crash in TEntryList

This is a follow up of f5249760 and a fix for ROOT-10632

Fix discrepancy with find_package_handle_standard_args and  name of the calling package (CMake 3.17)

While gtk sources are in ROOT source directory, one needs
to specify includes from ${CMAKE_SOURCE_DIR}.
This is not good, while dictionaty intend to be build only from
${CMAKE_BUILD_DIR}/include.

Was required in previous usage, now they are fully internal