The installation of Homebrew + GCC necessitates Xcode Command Line Tools, which will be installed if it is not present.
The installation of Homebrew + GCC is optional and can be deselected by the user.
This is not strictly needed since currently there is no submodule within a
submodule in Dynare. However, this is useful for the Enterprise
version (dseries has a matlab-fame-io submodule there). Moreover, this will
prevent bad surprises in the future should we introduce such a 2nd-level
submodule.
The detection of the slicot library failed because xerbla has been moved out of
the BLAS implementation shipped with Octave (probably since 5.1.0). This commit
fixes our fake implementation of mkoctfile so that it returns -lxerbla after
-lblas (as is done in the official Windows package for Octave).
Since 533a4ff7db, several MEX files were missing
in the Windows package (gensylv, k-order, dynare_simul, MS-SBVAR). The problem
was that the environment variable containing the directory containing MSYS2
libraries (incl. matio and gsl) was not correctly passed to parallelized
subshells.
Previously, if Blanchard-Kahn conditions were not satisfied, there was the
possibility an exception about NaN/Inf values in gy/gu was raised *before* the
exception about BK itself, leading to a confusing message. This commits moves
the test for BK before the test for NaN/Inf in gy/gu.
Closes: #674
In 98bdf76581, all XLSX files were removed in
favour of XLS, for compatibility with MATLAB R2009b.
But this broke the very same tests under Octave, since XLS support is not very
good there.
This commit implements a solution that works everywhere, by using XLSX whenever
possible, and XLS otherwise.
— use XLS instead of XLSX in testsuite datafiles
— use optimset instead of optimoptions
— use quadv/quadl/quadgk instead of integral
— fix race condition in load_m_file_data_legacy.m tests
— use fallback implementation for intersect(…, 'stable')
Previously, the solver would recompute the Jacobian at every iteration. But, if
an iteration fails, the current point is not moved (only the radius of the
trust region changes), hence it is not necessary to recompute the Jacobian in
that case. This commit implements this optimization.
* split up report output, creating new files for the preamble, the body of the report, and each individual page of the report.
* allow the user to set the page number
* allow for the removal of headers and footers from a given page