3.14.16

• Fixed load of linear solver libraries at runtime on Windows, which was broken for relative paths in 3.14.15 [#759, #760].

3.14.15

• Fixed include guard of IpGenAugSystemSolver.hpp [#756, by Christopher Wellons].
• Improved finding dependencies of linear solver libraries loaded at runtime on Windows [#755, by Yue Yang].

3.14.14

• Fixed build of MA57 interface when FUNNY_MA57_FINT is defined (MA57 with long integers).
• Fixed that initial dual values for fixed variables (z_L, z_U) were ignored when doing a warm start with fixed_variable_treatment = make_constraint [#728].
• Fixed that the last argument when calling MA27C[D] in MA27 interface was wrong [#738].

3.14.13

• Reduced priority for making Spral the default value for option linear_solver [#677].
• Adapted to change of integer types in Spral interface. Minimal required Spral version is now v2023.03.29.
• Fixed that return code from TNLP::eval_jac_g() was ignored at first call [#697, by Christoph Hansknecht].
• Print additional messages when reallocation of MA27 working space failed [#671, by Daniel Oliveira].
• Added option file_append to specify whether to append to output_file instead of truncating it.
  Added default argument file_append to Journalist::AddFileJournal(), added default argument fappend to FileJournal::Open(), and added default argument file_append to IpoptApplication::OpenOutputFile(). [#720]

3.14.12

• Fix that a source file was installed and install more header files. [#641, #642, by Joao Sousa Pinto]
• Fixed crash of GetIpoptCurrentIterate() and GetIpoptCurrentViolations() in C interface when called before or after IpoptSolve(). [#644, #645, by Robbybp]
• Updated HSL_MA97 header file to the one from HSL MA97 2.8.0 [#646, by Jari Fowkes].
• Fixed crash when trying to solve problem without variables and constraints. [#648]
• Added optional argument to AlgorithmBuilder constructor to provide name of custom solver. [#618]
• Changed handling of dual solution for square problems: When solving a problem with as many equations as variables, Ipopt used to ignore the violation of dual feasibility and complementarity in the convergence check and computed a final dual solution via a least-square estimate. If this failed, Ipopt would claim a square problem to be solved to optimality without providing a solution that satisfies optimality conditions. With this version, the behavior has been changed so that dual feasibility is no longer ignored by the convergence check, the least-square estimate is only computed if optimality is not proven already, and the normal Ipopt algorithm continues if the least-square estimate does not satisfy optimality conditions.
• Updated HSL_MC86 header file to the one from HSL MC68 3.3.3 [#653, by Jari Fowkes].

3.14.11

• Added IpoptData::TimingStats() const [#611]
• Assume DLL library extension in linear solver library loader on Windows also when building with other compiler than MSVC/Intel [#628].
• Updated buildsystem files after upgrading to most recent versions of autotools.
• Install some additional header files [#637].

3.14.10

• Added option grad_f_constant to specify that objective function is linear. If set, the gradient of the objective will be requested by Ipopt only once. [#597]
• Added OrigIpoptNLP::orig_d_L() and OrigIpoptNLP::orig_d_U() to get original constraint sides (before relaxation due to bound_relax_factor > 0).
• TNLP::get_curr_violations() now returns the constraint violation and complementarity with respect to the original (non-relaxed) constraint sides. [#603]

3.14.9

• Fixed mapping of meta data for variable bounds, e.g., variable names, from TNLP to Ipopts internal NLP [#590].

3.14.8

• Added options ma27_print_level, ma57_print_level, and mumps_print_level to enable output from these linear solvers.

For Windows libraries, use the one from the 3.14.7 release.

3.14.7

• Fixed that ComputeSensitivityMatrix() of sIpopt assumed that there are no more than two parameters [#578, by Andrea Vescovini].
• For completeness, added option gradient_approximation to enable approximation of gradient of objective function by finite differences. Do not use. [#573]
• Added function IPSETPROBLEMSCALING to Fortran interface to set problem scaling [#577, by Steven R. Hall]

In the attached binaries, MUMPS is now included as a separate DLL [#584].