At the end of a thread, we must first notify the main thread waiting on the
condition variable, then unlock the mutex. We must do these two operations in
that order, otherwise there is a possibility of having the main process
destroying the condition variable before the thread tries to notify it (if all
other threads terminate at the same time and bring the counter down to zero).
For that reason, we cannot use std::notify_all_at_thread_exit().

Bug introduced in commit 752a02a3.

GCC 5 supports C++14, but it is still not enough because it does not implement
N4387 that was retroactively applied to the standard in 2015 (see the "Notes"
and "Defect reports" sections of
https://en.cppreference.com/w/cpp/utility/tuple/tuple).

Another version is already under tests/trend_var/ directory.

Also fix extensions of source files.

Previously it was plain TeX.

This allows us to remove the test for pdfetex.

By the way, simplify Makefile rules.

The material of this document comes from the introduction to the library that
had been lost in the move away from CWEB (formerly tl/cc/main.web).

This file gives a good overview of the library. It has been adapted from TeX to
LaTeX. Also I fixed a mistake in the Faà di Bruno's tensor formula.

Forgotten in commit 73330880.

The former default was 2 threads, since Dynare++ was written at a time when
hyper-threading was being generalized (but multiple cores were not).

On Windows, this means that a POSIX threads implementation is no longer needed,
since C++11 threads are implemented using native Windows threads.

On GNU/Linux and macOS, POSIX threads are still used under the hood.

A new m4 macro (AX_CXX11_THREAD) is used to add the proper compilation
flags (instead of AX_PTHREAD).

Update dynare.texi

See merge request Dynare/dynare!1642

In particular, the test binary now errors out in case of test failure.

- Remove the GeneralMatrix(const ConstVector &) constructor, since it is hides
  a memory allocation (copying the ConstVector into a fresh Vector). This
  helped detecting and fixing several unneeded memory allocations. Some other
  memory allocations are now more visible (with an explicit Vector{}
  constructor).

- Add checks in GeneralMatrix(Vector, …) and ConstGeneralMatrix(ConstVector, …)
  constructors for verifying that the {Const,}Vector has unit-stride (this was
  an implicit assumption so far) and is large enough for storing rows*cols
  elements.

- Add GeneralMatrix::operator=(const ConstGeneralMatrix &).

- Delete ConstGeneralMatrix::operator=().

Many BLAS/LAPACK calls were making the assumption that LD==rows when passing
matrices. In some cases this was correct (because of implementation details,
in particular because how the copy constructor of GeneralMatrix is implemented).
But in other cases it was a bug.

With this commit, the actual value for LD is systematically used (this fixes
existing bugs and prevent possible future ones if the implementation details
were changed).

- these classes now encapsulate a std::shared_ptr<{const, }double>, so that
  they do not perform memory management, and several {Const,}Vector instances
  can transparently share the same underlying data

- make converting constructor from ConstVector to Vector explicit, since that
  entails memory allocation (but the reverse conversion is almost costless, so
  keep it implicit); do the same for GeneralMatrix/ConstGeneralMatrix,
  TwoDMatrix/ConstTwoDMatrix

- remove the constructors that were extracting a row/column from a matrix, and
  replace them by getRow() and getCol() methods on {Const,}GeneralMatrix

- rename and change the API of the complex version Vector::add(), so that it is
  explicit that it deals with complex numbers

- add constructors that take a MATLAB mxArray