Ipopt Documentation  
ChangeLog.md
Go to the documentation of this file.
1 # ChangeLog
2 
3 Here we list changes of Ipopt since the release of version 2.2.0.
4 More detailed information about incremental changes can be found in the
5 [commit history](https://github.com/coin-or/Ipopt/commits).
6 [TOC]
7 
8 ## 3.14
9 
10 ### 3.14.10 (2022-10-11)
11 
12 - Added option `grad_f_constant` to specify that objective function is linear.
13  If set, the gradient of the objective will be requested by Ipopt only once. [#597]
14 - Added `OrigIpoptNLP::orig_d_L()` and `OrigIpoptNLP::orig_d_U()` to get
15  original constraint sides (before relaxation due to bound_relax_factor > 0).
16 - `TNLP::get_curr_violations()` now returns the constraint violation and
17  complementarity w.r.t. the original (non-relaxed) constraint sides. [#603]
18 
19 ### 3.14.9 (2022-07-21)
20 
21 - Fixed mapping of meta data for variable bounds, e.g., variable names,
22  from TNLP to Ipopts internal NLP [#590].
23 
24 ### 3.14.8 (2022-07-13)
25 
26 - Added options ma27_print_level, ma57_print_level, and mumps_print_level
27  to enable output from these linear solvers.
28 
29 ### 3.14.7 (2022-06-24)
30 
31 - Fixed that ComputeSensitivityMatrix() of sIpopt assumed that there are
32  no more than two parameters [#578, by Andrea Vescovini].
33 - For completeness, added option `gradient_approximation` to enable approximation
34  of gradient of objective function by finite differences. Do not use. [#573]
35 - Added function `IPSETPROBLEMSCALING` to Fortran interface to set problem
36  scaling [#577, by Steven R. Hall]
37 
38 ### 3.14.6 (2022-05-02)
39 
40 - Fixed mapping of meta data for inequalities, e.g., constraint names,
41  from TNLP to Ipopts internal NLP [#570, by Daniel Oliveira].
42 - Fixed that MC68 ordering time was not accounted in symbolic factorization
43  time of HSL MA86 [#571].
44 - Include more header files in IpIpoptCalculatedQuantities.hpp for setups
45  where forward declarations are not sufficients [#572].
46 
47 ### 3.14.5 (2022-02-09)
48 
49 - Tried to fix recognition of JNI headers on macOS >= 11 [#516].
50 - Fixed that only primal variable values where passed to `finalize_solution()`
51  when a timelimit was reached [#552].
52 
53 ### 3.14.4 (2021-09-20)
54 
55 - Skip build of Java interface if either java or jar is not found [#510].
56  Only give warning if javac and jar are found, but no java or javadoc.
57 - Fixed that `--with-lapack-lflags` was ignored if `--with-lapack` was not
58  specified explicitly [#512,#515].
59 
60 ### 3.14.3 (2021-09-03)
61 
62 - Fixed timing for iterate initialization if initialization fails due to
63  an evaluation error.
64 - Fixed possible integer overflow when reserving space for indices of Jacobian
65  belonging to fixed variables (introduced with 3.14.0) and reduced memory
66  usage for indices of Jacobian belonging to fixed variables.
67 
68 ### 3.14.2 (2021-07-21)
69 
70 - Added `OptionsList::UnsetValue()` to remove an option setting.
71 - Added missing translation of some Ipopt status codes into AMPL solve result codes.
72 - If using the MPI-parallel version of MUMPS: Moved calls to
73  `MPI_Init()`/`MPI_Finalize()` in MUMPS interface into global constructor/destructor
74  of Ipopt library (if building with GCC/clang). Use configure flag
75  `--disable-mpiinit` to disable. [#500]
76 
77 ### 3.14.1 (2021-06-25)
78 
79 - Fixed invalidation of cached Hessians when reoptimizing with same structure.
80 - Added `OptionsList::SetBoolValue()` and `OptionsList::SetBoolValueIfUnset()`. [#492]
81 - Skip check for and link against libdl if linear solver loader is disabled.
82 - Fixed missing initialization of `BacktrackingLineSearch::in_watchdog_`.
83 - Fixed a problem with the current solution not being reset when
84  initialization of a NLP fails in reoptimization.
85 - Fixed that C++11 capability was not correctly identified with MS/Intel compilers.
86 
87 ### 3.14.0 (2021-06-15)
88 
89 #### Data Types
90 
91 - Due to a contribution by Mitchell Clement [#428], it is now possible
92  to build Ipopt in a variant that uses single-precision floating
93  point arithmetic instead of the default double-precision.
94  This can be enabled by specifying the configure flag `--with-precision=single`.
95  Doing so has a number of consequences on Ipopt interfaces and Ipopt dependencies.
96  See the Ipopt installation instructions for more details.
97 - The name of all functions in `IpBlas.hpp` and `IpLapack.hpp` has been
98  changed to drop the precision-specifier "D" in the name.
99  Wrapper using the old names are available, but their use is deprecated.
100 - It is now possible to build Ipopt in a variant that uses 64-bit
101  integers for its `Index` type. This can be enabled by specifying
102  the configure flag `--with-intsize=64`. Doing so has a number of
103  consequences on Ipopt interfaces and Ipopt dependencies. See
104  the Ipopt installation instructions for more details. [#259]
105 - Added new header `IpTypes.h` that defines Ipopt types for integer
106  and real numbers for C users.
107 - Deprecated almost never used type `Ipopt::Int`, use `int` instead.
108 - Deprecated `Number`, `Index`, and `Int` (on global namespace) in
109  `IpStdCInterface.h`, use `ipnumber`, `ipindex`, and `int` instead, respectively.
110 - Deprecated `Bool` in `IpStdCInterface.h` and replaced it by `bool`
111  from `stdbool.h`. Note, that `Bool` was defined to be `int`, but `bool`
112  is likely a shorter type, e.g., `signed char`. Deprecated `TRUE` and
113  `FALSE`, use `true` and `false` instead.
114 - Deprecated `IPOPT_EXPORT` macro and introduced `IPOPT_CALLCONV`.
115 - Deprecated `IPOPT_FORTRAN_INTEGER_TYPE` and `ipfint`. These were
116  always assumed to be `int`. Use `ipindex` or `ipopt::Index` instead.
117 
118 #### Linear Solver Interfaces
119 
120 - Due to a contribution by Byron Tasseff [#446], it is now possible to use
121  the linear solver [SPRAL](https://github.com/ralna/spral) (Sparse Parallel
122  Robust Algorithms Library) with Ipopt.
123  SPRAL is open-source and can, optionally, make use of NVIDIA GPUs.
124  If Ipopt has been build with SPRAL, then option `linear_solver` can be
125  set to `spral` to enable use of SPRAL.
126  See the installation instructions on how to build the Ipopt/SPRAL interface
127  and the options documentation for new options that are available for
128  the Ipopt/SPRAL interface.
129 - Added `IpoptLinearSolvers.h` with function `IpoptGetAvailableLinearSolvers()`
130  to retrieve information which linear solvers are available for Ipopt
131  (linked in or loaded at runtime). Options `linear_solver` and
132  `linear_system_scaling` can now only be set to values which corresponding
133  code is available (linked in or for load at runtime).
134 - Revised and streamlined implementation of feature that loads libraries
135  with HSL or Pardiso routines at runtime. Added options `hsllib` and `pardisolib`
136  to specify name of of HSL and Pardiso libraries, respectively.
137 - Separated interfaces for Pardiso from [pardiso-project.org](https://pardiso-project.org)
138  and Pardiso from Intel MKL and allow to have both Pardiso versions available
139  with the same Ipopt libraries.
140  - Pardiso from pardiso-project.org can be selected via `linear_solver=pardiso`.
141  configure option <b>`--with-pardiso` should now specify only a Pardiso
142  library to be loaded at runtime</b> (the value for `--with-pardiso` decides
143  the default value for option `pardisolib`).
144  To avoid conflicts with Pardiso from MKL, it is **no longer possible to
145  link against Pardiso from pardiso-project.org**.
146  - Pardiso from MKL can be selected via <b>`linear_solver=pardisomkl`</b>.
147  - Options that influence Pardiso from MKL are named <b>`pardisomkl_msglvl`, `pardisomkl_order`, etc.</b>
148  - configure option `--disable-pardisomkl` can be used to disable the check
149  for MKL Pardiso. [#454]
150 - Fixed that return code (`info%flag`) of `ma97_solve()` was not checked
151  in `Ma97SolverInterface::MultiSolve()`.
152 - Calls into MUMPS are now protected by a mutex if compiling for C++11 or higher.
153  This prevents Ipopt from calling MUMPS concurrently from several threads.
154 - An insufficient memory return status is now also given if the memory
155  required for a working space array of a linear solver exceeds the maximal
156  value for the working space length variable, e.g., if MA27 requires a
157  working space array of length higher than 2^31.
158 - Avoid floating point overflow when computing memory increase in interfaces
159  to MA27 and MA57.
160 
161 #### Algorithm
162 
163 - Changed treatment of NLPs with all variables fixed and `fixed_variable_treatment`
164  set to `make_parameter`:
165  Ipopt will still terminate early, but initialize its data structures
166  first and print a solve summary (49a0d3a56).
167  Changed treatment of NLPs with inconsistent variable bounds or constraint
168  sides: These will now result in an invalid problem definition error (-11)
169  (5cdb2624bb).
170  As a consequence, solve statistics should now always be available if the
171  return status from `(Re)Optimize()` is larger than -10.
172 - Bound multipliers are now computed for fixed variables if
173  `fixed_variable_treatment` is set to `make_parameter` (the default).
174  This can trigger a reevaluating of the gradient of the objective
175  function or the Jacobian of the constraint functions. If this
176  is not desired, then option `fixed_variable_treatment` can be set
177  to the new value `make_parameter_nodual`. [#308]
178 - Changed default for `honor_original_bounds` to no.
179  Variable bounds (see option `bound_relax_factor`) are now relaxed
180  by at most the value of `constr_viol_tol`.
181  The solve summary now reports the violation of original bounds
182  and `SolveStatistics::(Scaled)Infeasibilities()` has been extended
183  to report the violation of original variable bounds. [#312]
184 - If Ipopt hits a time or iteration limit during watchdog phase,
185  the iterate from before the watchdog phase is now restored and
186  passed to `finalize_solution`. Note that this does not apply if
187  Ipopt is stopped due to a user-interrupt (`intermediate_callback`).
188  [#289]
189 - When `BacktrackingLinesearch` could not find a trial point that
190  provided sufficient progress, it may resort to call the feasibility
191  restoration phase. If, however, the current (scaled) constraint
192  violation was below tol/100, this would result in the infamous
193  "Restoration phase is called at almost feasible point..." abort
194  with status code `Restoration_Failure`. Now, this message has been
195  changed to "Linesearch failed, but no restoration phase or other
196  fall back is available." and the status code to `Error_In_Step_Computation`
197  to better reflect that the linesearch failed and not the restoration
198  phase. Further, the unscaled constraint violation tolerance now
199  needs to be below `constr_viol_tol/10` as well in order to trigger
200  this abort.
201 - When a square problem is solved and the restoration phase only succeeded to
202  find a point that is feasible w.r.t. constr_viol_tol, but not w.r.t. tol,
203  then status Feasible_Point_Found is returned now.
204 
205 #### Ipopt interfaces
206 
207 - Added `TNLP::get_curr_iterate()` and `TNLP::get_curr_violations()`
208  to request the current iterate (primal and dual variable values)
209  and primal and dual infeasibility w.r.t. the TNLP. The methods
210  are meant to be called during `intermediate_callback` to inspect
211  the current iterate. The C, Fortran, and Java interfaces have
212  been extended by corresponding functions, too. The hs071 examples
213  have been extended to show use of the new functions. Added test
214  `getcurr` to test new functions. [#382, #451]
215 - Extended the Java interface by the possibility to specify an `intermediate_callback`.
216 - Added flag to `TNLPAdapter::ResortX()` to specify how to handle fixed variables.
217  Added flag to `TNLPAdapter::ResortG()` to specify whether to correct by
218  right-hand-side of equality constraints.
219  Added `TNLPAdapter::ResortBoundMultipliers()` to generate correct duals
220  for fixed variables if `fixed_variable_treatment` is `make_constraint`.
221  Added `TNLPAdapter::GetFullDimensions()`, `TNLPAdapter::GetFixedVariables()`,
222  `TNLPAdapter::GetPermutationMatrices()`, and `TNLPAdapter::GetC_Rhs()`
223  to retrieve more information of the transformation in a `TNLPAdapter`.
224 - Added `TNLPAdapter::ResortBounds()` and deprecated `TNLPAdapter::ResortBnds()`.
225 - `AmplTNLP` constructor and `AmplTNLP::get_options()` now expect to
226  receive a pointer to a `RegisteredOptions` object as well.
227  Previous versions of these methods are still available but deprecated.
228  Using a NULL pointer for the `RegisteredOptions` argument is
229  also deprecated.
230 - `std::overflow_error` exceptions are now caught by Ipopt even if rethrowing
231  of non-Ipopt exceptions is enabled.
232 
233 #### Timing
234 
235 - Added option `max_wall_time` to specify a wallclock time limit.
236  Added `SolverReturn` code `WALLTIME_EXCEEDED` and `ApplicationReturnStatus`
237  code `Maximum_WallTime_Exceeded`.
238 - Changed default for `max_cpu_time` to 1e20 to indicate no CPU timelimit.
239 - Detailed timing statistics are no longer collected by default.
240  To reenable, set the new option `timing_statistics` to yes or set the
241  option `print_timing_statistics` to yes [#299].
242 - Removed `IpoptData::ResetCpuStartTime()`. Deprecated `IpoptData::cpu_time_start()`,
243  use `IpoptData::TimingStats()::OverallAlgorithm()::StartCpuTime()` instead.
244 - Deprecated `SolveStatistics::TotalCPUTime()`, use `SolveStatistics::TotalCpuTime()`
245  instead.
246 
247 #### Option handling
248 
249 - All Ipopt options are now also available via the AMPL interface,
250  that is, can be set in AMPL via an option `ipopt_options "..."` statement.
251 - Added flag "advanced" for `RegisteredOption` to store whether an
252  option is rather meant for expert users.
253 - Added class `RegisteredCategory` to store information on category
254  of registered option. Next to the name, this is a priority, which
255  determines the order in which categories are printed. As a
256  consequence, methods `RegisteredOption::RegisteringCategory()` and
257  `RegisteredOptions::RegisteringCategory()` now return a
258  `RegisteredCategory` object instead of a string. Further,
259  `RegisteredOption::SetRegisteringCategory()` has been removed.
260 - Deprecated existing `RegisteredOptions::OutputXyz()` methods and
261  added new variant of `RegisteredOptions::OutputOptionDocumentation()`.
262  Added parameter `print_advanced_options` to indicate whether to
263  print documentation for advanced options, too.
264 - Added previously undocumented advanced options to options reference.
265 
266 #### Miscellaneous
267 
268 - Updated buildsystem, including improved checks for dependencies,
269  use of current autotools, and skipping the build of intermediate
270  non-distributed libraries.
271 - Fixed missing initialization of `IpoptApplication::smart_jnlst` in
272  second `IpoptApplication` constructor, if `ipopt verbosity > 0`.
273 - Add GCC format attribute to `Journalist::Printf` functions to enable
274  `printf`-formatter check.
275 - Fixed `DenseVector::SumLogsImpl()` such that it returns 0.0 for
276  a vector of dimension 0. Returned `nan` for homogeneous 0-vector
277  of dimension 0 before, which may have caused the restoration phase to
278  fail for problems with only equality or only inequality constraints.
279  Also other `DenseVector` methods now skip calculations when dimension
280  is 0 to avoid (probably harmless) divisions by zero.
281 - Fixed a problem where moving slack away from 0 did not succeed
282  when mu was very small. [#212]
283 - Fixed a problem where moving slacks away from 0 resulted in `nan`
284  if multipliers were zero. Added `Vector::ElementWiseSelect()`.
285 - Various tiny bugfixes and improvements in performance and code style
286  by following suggestions of `cppcheck`.
287 - Added documentation on some available C preprocessor flags for expert users.
288 - Fixed static build of sIpopt without GCC. Fixed that installed sIpopt
289  headers were not usable (`SIPOPTLIB_EXPORT` not defined).
290 - Fixed wrong gradient of objective function and Lagrangian Hessian in
291  restoration problem [#478, by Nai-Yuan Chiang].
292 - If Ipopt is compiled for checklevel 2 or higher and the GLIBC extension `feenableexcept()`
293  is available, then floating-pointing exceptions divbyzero, overflow, and
294  invalid are raised while `IpoptAlgorithm::Optimize()` is running.
295 - Fixed that norm on unscaled complementarity or scaled complementarity tolerance
296  were negative when maximizing (by using a negative scaling factor for the
297  objective).
298 - Changed formula for relative error in derivative checker. The absolute
299  error is now scaled up if the approximate derivative value is between
300  derivative_test_tol and 1. [#487].
301 - The second-order derivative checker now uses values for obj_factor and
302  lambda that are different from 1.
303 
304 
305 ## 3.13
306 
307 ### 3.13.4 (2021-02-24)
308 
309 - Fixed a linking issue for `ipopt_sens` [#418]
310 - Fixed Makefile for Java example regarding location of jar file
311 - Fixed build of R interface if using `-fvisibility=hidden`.
312 
313 ### 3.13.3 (2020-10-16)
314 
315 - Members of `AmplTNLP` class are now protected instead of private.
316 - Updated Eclipse Public License from 1.0 to 2.0.
317 - Fixed dangling pointer problems with Journalist used for debugging
318  (`--with-ipopt-verbosity > 0`) when more than one `IpoptApplication`
319  is used. [#393, thanks to Brad Bell]
320 - Fixed build problem when using HSL library that does not include
321  MA27, MA57, or MC19. [#395]
322 - Added example `recursive_nlp` that uses Ipopt to solves an
323  optimization problem for the evaluation of the objective function.
324  [contributed by Brad Bell]
325 - Fixed build of linear-solver loader on Windows [#408]
326 
327 ### 3.13.2 (2020-04-30)
328 
329 - The C-preprocessor defines `COIN_IPOPT_CHECKLEVEL`,
330  `COIN_IPOPT_VERBOSITY`, and `FORTRAN_INTEGER_TYPE`, which are defined
331  by `IpoptConfig.h`, have been renamed to `IPOPT_CHECKLEVEL`,
332  `IPOPT_VERBOSITY`, and `IPOPT_FORTRAN_INTEGER_TYPE`, respectively.
333  They are still available under their previous name, but these
334  will be removed in Ipopt 3.14.
335 - Changed dependencies as used by coinbrew to use new versions (2.1)
336  of ThirdParty/HSL and ThirdParty/MUMPS and dropped ThirdParty/Metis.
337  The new versions of the HSL and MUMPS build scripts now look
338  for a Metis library in the system and should work with both
339  Metis 4 and Metis 5.
340 - Changed location where Java interface jar gets installed from
341  `$libdir` to `$datadir/java/`.
342 - minor fixes to buildsystem
343 
344 ### 3.13.1 (2020-03-11)
345 
346 - Added asserts that check whether sparsity pattern of Jacobian
347  and Hessian as returned by TNLP are within range w.r.t. number
348  of variables and constraints. [#350]
349 - `TNLPAdapter::ResortBnds` now initializes complete output arrays
350  with 0.0 before filling in values for non-fixed variables. Use
351  new argument `clearorig` to turn this off. [#352]
352 - bring back configure variables `ADD_{C,CXX,F}FLAGS`
353 - added configure option `--enable-relocatable` to make prefix in
354  pkg-config files relative to pcfiledir (assuming that `--libdir`
355  hasn't been set)
356 - bring back `configall_system.h` for build without config header
357 - minor fixes to buildsystem
358 
359 ### 3.13.0 (2019-10-19)
360 
361 This major release comes with a larger renovation of the build
362 system and a changed directory structure (eliminated top directory),
363 which is the result of a long and still on-going effort to use
364 recent autotools versions for various COIN-OR projects, reduce
365 future maintenance efforts, and adapting behaviors of standard
366 autotools-based projects.
367 As a consequence, a monolithic build of Ipopt, which builds Ipopt
368 with all its dependencies in one run of configure and make is no
369 longer possible. Dependencies should now be build and installed
370 before building Ipopt.
371 Additionally, support for some outdated versions of dependencies
372 and unmaintained components of Ipopt has been dropped and some
373 improvements that may require changes on the users side have been
374 applied.
375 
376 A more detailed, probably incomplete, list of changes follows:
377 - Removed git submodules. Dependencies (HSL, Mumps, ASL, etc) now
378  need to be build and installed in advance, either manually or
379  by using coinbrew.
380 - Dropped support for HSL < 2013.
381 - Dropped support for MA28 in the linear solver loader.
382 - Dropped support for Pardiso < 4.0 from pardiso-project.org.
383 - Added support for Mumps 5.2.x, though initial experiments on
384  CUTEst indicated that, on average, performance is worse than
385  when using Mumps 4.10.0.
386 - Dropped CUTEr interface, the successor CUTEst includes an
387  interface to Ipopt.
388 - Dropped Matlab interface as it is unmaintained and it was
389  reported that it stopped functioning.
390  Use https://github.com/ebertolazzi/mexIPOPT instead.
391 - Dropped MSVS project files as unmaintained and not functioning
392  with current Ipopt anymore.
393 - Integrated Java interface into the main Ipopt library, that is,
394  it is handled equivalently to the C and Fortran interfaces:
395  - The source moved into `src/Interfaces`.
396  - The JNI functions are now included in the main Ipopt library,
397  thus an extra jipopt library is no longer build or necessary.
398  - The Java class and `org.coinor.ipopt.jar` package are build and
399  installed as part of the main Ipopt build.
400  - The examples moved into `examples/*_java`.
401  - A Java interface test is executed by `make test`.
402  - To build javadoc, run `make javadoc` in the main build directory.
403  - The configure flag `--disable-java` can be used to disable the
404  check for Java and build of the Java interface.
405  - `DLLPATH` and `DLLNAME` have been removed from the Ipopt class and
406  constructors that works without arguments and with only one
407  argument (specifying the Ipopt library namestem) have been added.
408  - Method `Ipopt::finalize` has been marked as deprecated and will
409  be removed in some future Ipopt version. Users must call
410  `dispose()` explicitly.
411 - Integrated sIpopt into the main Ipopt build, that is, it is now
412  build together with Ipopt, but installed as separate library
413  and executable. Use `--disable-sipopt` to disable building sIpopt.
414 - `IPOPT_THREAD_LOCAL` now uses C++11's `thread_local` keyword if C++11
415  is available.
416 - When using a GCC-compatible compiler, Ipopt and sIpopt interface
417  functions are now declared with `visibility(default)`-attribute,
418  thus building Ipopt with `-fvisibility=hidden` still produces a
419  usable library.
420 - When using a MSVC-compatible compiler, Ipopt and sIpopt interface
421  functions are now declared with `dllimport`-attribute, so that an
422  Ipopt C++ DLL can be used.
423 - Under Windows/Msys2, DLLs are now build by default.
424 - Cygwin and MSys1 are not supported.
425 - pkg-config is now mandatory to use dependencies like ASL or HSL.
426  On Windows, make sure to use a pkg-config version that produces
427  Unix-style paths.
428 - Script "`compile`" is now used to wrap around calls of cl/icl/ifort
429  and translate GCC-style compiler flags to MSVC style.
430 - "Addlibs" files have been removed, pkg-config should be used instead.
431 - Header files are now installed in the better named
432  `$prefix/include/coin-or` instead of `$prefix/include/coin`.
433 - The default for `--prefix` is no longer the build directory, but
434  the autotools-default, probably `/usr/local`.
435 - The check for a Fortran compiler can be disabled via `--disable-f77`
436  and Ipopt can easier be build without a Fortran compiler.
437 - Lapack is no longer optional, but required. The separate check
438  for Blas and the `--with-blas` flags have been removed.
439 - `--enable-debug` does not imply `--disable-shared` anymore.
440 - Removed `--enable-debug-ipopt`, use `--enable-debug` instead.
441 - Removed configure variables `{ADD,OPT,DBG}_{C,CXX,F77}FLAGS`.
442  Use `{C,CXX,F77}FLAGS` instead.
443 - Silent build output is now enabled by default, use configure
444  flag `--disable-silent-rules` or call make with `V=1` to disable.
445 - Also for static builds, PIC objects are now generated by default,
446  use `--without-pic` to disable.
447 - The `--with-*-incdir` and `--with-*-lib` configure flags have been
448  replaced by corresponding `--with-*-cflags` and `--with-*-lflags`
449  flags. Note that the include directories need to be specified
450  via `-I<dir>` in `--with-*-cflags`.
451 - Fixed handling of `ma77_default_control` in `LSL_setMA77()`.
452 - Fixed calculation of quality function when setting option
453  `quality_function_centrality` to `reciprocal`.
454 - Fixed compiler warnings, in particular when using `-Wunused-parameter`.
455 - Changed default for `ma97_print_level` to -1. This avoids messages
456  about numerical singular systems written to stdout by default.
457 
458 ## 3.12
459 
460 ### 3.12.13 (2019-04-08)
461 
462 - fixed Pardiso settings when using Pardiso from Pardiso project
463  website (by Olaf Schenk): the new settings should provide much
464  better performance; the default for option `pardiso_order` changed
465  from `five` to `metis`.
466 - changed distinction of MKL and Basel Pardiso in configure: to
467  use MKL Pardiso, only specify MKL for Blas; to use Basel Pardiso,
468  use `--with-pardiso`
469 
470 ### 3.12.12 (2018-11-17)
471 
472 - allow for `--without-matlab-home` to disable check for Matlab [r2748]
473 - add `dppsv` to `v8-ifort` [r2746]
474 - disable error in `LibraryHandler.c` if `snprintf` detection failed [r2751]
475 
476 ### 3.12.11 (2018-09-16)
477 
478 - fill MUMPS struct with zeros when allocating in MUMPS interface [r2724]
479 - minor fix in build-system of ThirdParty/ASL
480 
481 ### 3.12.10 (2018-06-02)
482 
483 - fixed setting for parallel solve when using MKL Pardiso
484  (by t1393988511) [r2711]: parallel solve was disabled (which
485  is not the default); note, that the setting for parallel
486  factorization was not affected
487 - fixed invalid read in AMPL interface for problems without
488  objective function [r2715, #305]
489 - updated ThirdParty/ASL to retrieve updated ASL (20180528) [#305]
490 - name JIpopt library `libjipopt.dylib` on Mac OS X [r2718, #275]
491 
492 ### 3.12.9 (2018-01-15)
493 
494 - fixed memory leak in MA86 interface (by mhahn) [r2700,#283]
495 - fixed handling of time limit when reoptimizing: CPU time spend
496  was accumulated when reoptimizing, while it should have been
497  reset for each solve (by paul-scott) [r2702,r2703]
498 - fixed sign in Jacobian finite-difference approximation when point
499  was close to variable upper bounds (by Enrico Bertolazzi) [r2704]
500 
501 ### 3.12.8 (2017-06-12)
502 
503 - add define for `FORTRAN_INTEGER_TYPE` to `config_ipopt_default.h`
504 - `IpoptApplication::RethrowNonIpoptException()` now returns whether
505  non-ipopt exceptions were rethrown before the method was called.
506 
507 ### 3.12.7 (2017-02-25)
508 
509 - removed compiler flag `-pedantic-errors` to avoid problems with some
510  configure tests when using recent GCC versions
511 - fixed rare bug in handling variable/constraint names in `AmplTNLP`
512  (by G. Hackebeil) [r2673]
513 - the `get.Mumps` script in ThirdParty/Mumps now renames `libseq/mpi.h`
514  to `libseq/mumps_mpi.h` to avoid conflicts when building in a MPI
515  environment (by T. Ralphs); note that if updating an existing
516  checkout/download of Ipopt, you may have to rerun get.Mumps
517 
518 ### 3.12.6 (2016-07-20)
519 
520 - better support for custom algorithm development [r2659] (by Gabriel Hackebeil):
521  "Reorganization of the `AlgorithmBuilder` class to allow easier
522  customization of the Ipopt algorithm. In particular, we wanted to
523  make use of the code that creates the `SymLinearSolver` object to
524  implement our own `SymLinearSolver` without copy-pasting everything
525  in AlgorithmBuilder.
526  `AlgorithmBuilder::BuildBasicAlgorithm` now consists of 8 method calls
527  that build the core components passed into the arguments of the
528  `IpoptAlgorithm` class. These calls are ordered based on any dependencies
529  they might have. In addition, all code for creating the `PDSystemSolver`,
530  `AugSystemSolver`, and `SymLinearSolver` has been moved into separate factory
531  methods.
532  Also, included is a change to install a few more header files with Ipopt.
533  Some of these are required to subclass AlgorithmBuilder, and the others
534  are simply some matrix types that we require."
535 - extend build system to work without Fortran compiler [r2660,r2661]:
536  If no Fortran compiler is available (`F77=unavailable`), then
537  the build system checks for functions in Blas, Lapack, and
538  Pardiso via C linkage. This seems to work when using the Intel MKL,
539  thus allowing to build Ipopt with C/C++ compilers and MKL only.
540  The linear solver loader and the CuteR interface are disabled when
541  no Fortran compiler is available. A user may have to adjust the
542  definition of `F77_FUNC` in `Ipopt/src/Common/IpoptConfig.h`.
543 
544 ### 3.12.5 (2016-04-30)
545 
546 - changed `fptr` from `long` to `void*`: the Fortran side needs to
547  make sure that it uses a big enough integer type to store a
548  C pointer, thus `void*` can be used on the C side [r2599]
549 - added additional second-order-correction method, which can be
550  selected by setting the new option `soc_method` to 1 (by Wei Wan)
551  [r2606, r2607]
552 - added parameter `allow_clobber` with default value false to
553  `IpoptApplication::Initialize()` and `OptionsList::ReadFromStream()`
554 
555 ### 3.12.4 (2015-08-09)
556 
557 - option to use regularized Hessian when doing a curvature test
558  without inertia information (`neg_curv_test_tol` > 0), new
559  option `neg_curv_test_reg` to switch back to original behavior
560  (by N.-Y. Chiang and V. Zavala Tejeda) [r2579]
561 - sIpopt: Added access to sensitivity directional derivative
562  vector (`ds/dp*(p-p0)` Eq. 14 sIpopt implementation paper). Also,
563  added an option to compute the sensitivity matrix and provide
564  access to it. Finally, added an example that shows how to
565  access the new information. (by R. Lopez-Negrete)
566 - use workaround for failing check for random number generator
567  with any gcc 4.8.x, x >= 2
568 
569 ### 3.12.3 and 3.11.11 (2015-04-15)
570 
571 - fixed bug in MA97 interface that lead to conversion issues
572  (by J. Hogg) [r2566, #260]
573 
574 ### 3.12.2 (2015-04-04)
575 
576 - revised integration of doxygen-generated documentation into build system
577  (by T. Ralphs)
578 
579 ### 3.12.1 (2015-02-13)
580 
581 - fixes to build system for dependency linking and library versioning
582 - Ipopt will now report an NLP with inconsistent variable bounds
583  or inconsistent constraints sides as infeasible instead of
584  throwing an invalid TNLP exception (by T. Kelman) [r2548]
585 
586 ### 3.12.0 (2015-01-23)
587 
588 - Library dependencies are now recorded in shared library builds,
589  which is intended to simplify linking against the Ipopt library.
590  However, the pkg-config and `ipopt_addlibs` files do not reflect
591  this change yet (it is rather experimental, imho). To restore
592  the previous behavior, use `--disable-dependency-linking` as
593  configure option.
594 - If linking against Intel MKL for Blas/lapack, use of Pardiso
595  from MKL is now automatically enabled. Note, that this will
596  change the default solver on Ipopt builds without any of the
597  linear solvers MA27, MA57, MA97, and MA86 (these take preference
598  over Pardiso). [#216]
599 - dropped support for old HSL sources (<2013; ThirdParty/HSLold)
600 - updated ASL sources, now downloaded from AMPL-MP (github)
601 - some internal changes to data structures (improved use of compound
602  component spaces) and addition of `IpLapackDppsv` (by Gabe Hackebeil)
603 
604 ## 3.11
605 
606 ### 3.11.10 (2015-01-18)
607 
608 - fix a memory allocation in Java interface in cases where `jint`
609  has a different size than `int` [r2513]
610 - the buildsystem now tries the `Accelerate` framework instead of
611  `vecLib` for finding Blas/Lapack on MacOS X
612 
613 ### 3.11.9 (2014-08-16)
614 
615 - fix compilation issue of Java interface on systems where `Index`
616  and `jint` have different size [r2498, #241]
617 - work around failing check for random number generator with gcc
618  4.8.3 [r2495, r2496]
619 - readded `IpTaggedObject.cpp` to list of sources to compile in
620  MSVS `v8-ifort` project file [r2492]
621 - work around missing support for thread-local storage with gcc < 4.5
622  on MacOS X [r2491, #243]
623 - fix call to MKL Pardiso init function [r2489]
624 - speed up Triplet to CSR converter [r2487, #234]
625 - fixed a bug in equilibration scaling where average values were
626  computed incorrectly (by V. Zverovich) [r2483]
627 
628 ### 3.11.8 (2014-04-08)
629 
630 - fixed a bug, introduced with Ipopt 3.11.0, where the tag in the
631  Ipopt's caching mechanism was not unique over time, which lead
632  to failures of Ipopt that were difficult to debug or recognize
633  (e.g., Ipopt may have stopped with an restoration failure for
634  instances that solved fine with Ipopt 3.10) [r2472, r2473]
635  I'm very thankful to Gabriel Hackebeil and Kurt Majewski for
636  their debugging effort on this issue.
637 - building Mumps with pthreads is now disabled by default [#229]
638 - fixed setting of `LD` on Windows (now set to link only iff using
639  MS/Intel compilers) [#230]
640 - fixed download link for Gnumex [r2471]
641 - for some messages about too-few-degrees-of-freedom and restoration
642  failure, the message level changed from error to strong-warning
643  [r2460, r2469]
644 - revised calls to `MPI_Init` and `MPI_Finalize` in MUMPS interface [r2467]
645  (`MPI_Init` is now called only if function `MPI_Initialized` is available
646  and MPI has not been initialized already; `MPI_Finalize` is only called
647  if Ipopt also called `MPI_Init`; ...)
648 
649 ### 3.11.7 (2013-12-18)
650 
651 - adapted PARDISO parameters when using MKL PARDISO to be close
652  to using Basel PARDISO
653 - added options `pardiso_max_iterative_refinement_steps` and
654  `pardiso_order`; the former defaults to 1 in case of MKL PARDISO,
655  which may help on instances that otherwise fail due to numerical issues
656 - removed duplicate code in `IpQualityFunctionMuOracle.cpp` [#225, r2445]
657 - fixed bug in triplet to csr converter [#226, r2446]
658 - minor changes in buildsystem
659 
660 ### 3.11.6 (2013-11-16)
661 
662 - updates to Matlab Interface build system (by T. Kelman)
663 - fix to updates of R Interface [r2416, #223]
664 - fixed `SHAREDLIBEXT` in `v8-ifort`'s `config.h` [r2426, #224]
665 - minor fixes to the buildsystem
666 
667 ### 3.11.5 (2013-10-26)
668 
669 - added method `IpoptApplication::RethrowNonIpoptException()` to enable
670  rethrowing of non-ipopt and non-`bad_alloc` exceptions catched in
671  the `*Optimize()` and `Initialization()` methods; default is still to
672  return with `NonIpopt_Exception_Thrown` status
673 - minor fixes to the buildsystem [#215, #222]
674 
675 ### 3.11.4 (2013-09-12)
676 
677 - hopefully fixed non-working linear solver loader in DLLs build with
678  MSVS/`v8-ifort` project files [r2365]
679 - allow MC19 to be loaded via linear solver loader (by J. Currie) [r2366]
680 - fixed new point flag when running dependency detector [r2368]
681 - experimental: adapt Pardiso interface to work with MKL Pardiso
682  (by J. Currie, T. Kelman) [r2369, #216]:
683  - in a few tests it has been found that Pardiso from Intel MKL nowadays
684  seems to work fine with Ipopt
685  - to use Intel MKL with Ipopt 3.11, one has to specify the MKL libs via
686  `--with-pardiso` and add `-DHAVE_PARDISO_MKL -DHAVE_PARDISO_PARALLEL`
687  to the compiler flags
688  - note that this is still an experimental feature (and thus not enabled
689  by default)
690 - updated Ipopt/R interface to version 0.8.4 [r2373]
691  - additional variables have been included in the object returned from `ipoptr`:
692  - `z_L`: final values for the lower bound multipliers
693  - `z_U`: final values for the upper bound multipliers
694  - `constraints`: final values for the constraints
695  - `lambda`: final values for the Lagrange multipliers
696  - removed `ipoptr_environment` as argument in `ipoptr` (see also r2372)
697 - fixed bug in penalty term for centrality in quality function (not used by
698  default) [#219, r2374]
699 - minor bugfixes in AMPL interface, debug print statements, and compound matrix
700  (by G. Hackebeil) [#218, r2371, r2377, r2378, r2379]
701 - download scripts for ASL, Blas, and Lapack now first try to download tarball
702  copies from the COIN-OR server
703 
704 ### 3.11.3 (2013-08-08)
705 
706 - `get.*` scripts for ThirdParty/{ASL,Blas,Lapack} now work around broken
707  ftp access to www.netlib.org.
708 
709 ### 3.11.2 (2013-07-01)
710 
711 - changed default for option `option_file_name` to `ipopt.opt`; specifying an
712  empty string for this option now disables reading of an option file [r2339]
713 - missing initial values are now set to 0.0, projected onto variable bounds,
714  in AMPL interface [r2340, #205]
715 - fixed missing variable initialization in MA97 interface [r2341, #206]
716 
717 ### 3.11.1 (2013-06-14)
718 
719 - the setup for the `v8-ifort` MSVS project changed to use dynamic runtime
720  DLLs instead of static linking, which caused crashes in debug mode
721  (by M. Roelofs) [r2301]
722 - fixed memory leaks in Java Interface (by javier) [#200, r2312]
723 - updates and fixes to MA77 and MA87 interfaces, adding support of
724  HSL 2013 codes (by J. Hogg);
725  HSL 2012 still supported when compiled with Ipopt, but the linear solver
726  loader to dynamically load a HSL library at runtime now assumes HSL 2013
727 - added option `ma97_solve_blas3` (by J. Hogg) [r2329]
728 - changed default for option `ma27_meminc_factor` from 10.0 to 2.0 [r2330]
729 - fixed bug in `ipopt_auxdata` of MATLAB Interface related to `iterfunc` [r2325]
730 
731 ### 3.11.0 (2013-05-07)
732 
733 #### Miscellaneous
734 
735 - update and extension of Ipopt documentation
736 - updated build of doxygen-generated documentation to comply with other
737  COIN-OR projects
738 - localized global variables in `TaggedObject` and `RegisteredOption`,
739  so that Ipopt should now be threadsafe as long as Ipopt objects
740  (esp. `SmartPtr`'s) are not shared between threads and a threadsafe
741  linear solver is used (e.g., MA27) [#167]
742 - no more need for whitespace character at end of options file
743 - added options `print_frequency_iter` and `print_frequency_time` to regulate
744  which iteration summary lines should be printed [#161]
745 - function evaluation timings are now available in `OrigIpoptNLP` [#86]
746 - some fixes to uncommon issues with the Ipopt `SmartPtr` [#162]
747 
748 #### Linear Solver Interfaces
749 
750 - new build system for Harwell codes (ThirdParty/HSL), which requires
751  the coin-hsl archives from http://www.hsl.rl.ac.uk/ipopt/;
752  previously downloaded HSL sources can still be used by placing them
753  into ThirdParty/HSLold, but this option will be removed in a future
754  Ipopt version
755 - new interfaces for Harwell codes HSL_MA77, HSL_MA86, and HSL_MA97;
756  see http://www.hsl.rl.ac.uk/ipopt/ about help on when to use which solver;
757  especially MA57 and HSL_MA97 should be considered as replacement for MA27;
758  however, MA27 is still the default for now
759 - changed default of `ma57_automatic_scaling` to `no` (faster in general,
760  but for higher reliability, you may want to set this option to yes)
761 
762 #### Ipopt Interfaces
763 
764 - major improvements for building the MATLAB interface (see documentation)
765 - MATLAB interface now returns number of function evaluations, too
766 - the MA57 interface can now be used with the MA57 library that comes with
767  MATLAB (configure option `--enable-matlab-ma57`; cannot use Metis)
768 - `auxdata` is now handled by a wrapper function `ipopt_auxdata.m` instead
769  of internally within the MEX code (replace Matlab call to `ipopt` with
770  `ipopt_auxdata` if using auxiliary data in any callbacks) [r2282]
771 - exposed more intermediate Ipopt information to `iterfunc` callback [r2283]
772 
773 - fixes to JIpopt buildsystem (now may work on windows and uses libtool)
774 - JIpopt now reads options file `ipopt.opt` by default, if present
775 - removed predefined `KEY_` strings in JIpopt
776 - renamed API functions that retrieve solution values in JIpopt
777 
778 - simplified installation of R interface
779 
780 ## 3.10
781 
782 ### 3.10.4 (2013-05-05)
783 
784 - fixed sign of dual values in AMPL solution again (with help of Gabe)
785  [r2169, r2170, r2184, #183]
786 - fixed bugs with reoptimizing a TNLP with all variables fixed [r2172, r2173, #179]
787 - fixed more issues with sparse data structures and non-double numbers
788  in Matlab interface (by T. Kelman) [r2191]
789 - added missing `final int` for Ipopt return code `Maximum_CpuTime_Exceeded`
790  in Java interface [r2216]
791 - fixed bug when trying to warmstart Ipopt in Java interface [r2253]
792 - fixed wrong use of `SmartPtr`'s in Java interface [r2255, r2263]
793 - fixed bug in returning final solution in Java interface [r2258]
794 - included patch in ThirdParty/Mumps to work around bugs in Mumps
795  matrix ordering routines AMF and QAMD (now give preference to AMD and METIS)
796 
797 ### 3.10.3 (2012-11-19)
798 
799 - minor fixes in MA86 interface (by Jonathan Hogg) [r2069, r2086]
800 - fix in `IpTripletToCSRConverter` for CSR forms with
801  extra entries (by Jonathan Hogg) [r2087]
802 - workaround problems with Mac OS X Lion's blas library
803  (by Frederic Hetch) [r2102, #181]
804 - the C interface now catches also Ipopt exceptions thrown
805  within the `OptimizeTNLP` call and returns `Ipopt::Unrecoverable_Exception`
806  as status [r2094, #144]
807 - fixed segmentation fault in adaptive barrier parameter update rule
808  when using the mehrotra option on unconstrained problems [r2114, #114]
809 - fixed segmentation fault in case no iterate is available in case of
810  catastrophic failure in restoration phase [r2115]
811 - fixed default for `mumps_dep_tol` to work with current Mumps versions [r2116]
812 - fixed sign of dual values in AMPL solution [r2123, #183]
813 - fixed issue with sparse gradients in Matlab interface
814  (by T. Kelman) [r2133, #187]
815 - sIPOPT (by H. Pirnay):
816  - starting values in C++ version of parametric example now
817  match AMPL version [r2098]
818  - numerical values in parametric example now match publication [r2099]
819  - added options `n_sens_steps` and `sens_boundcheck` as AMPL options [r2099]
820  - any non-zero suffix value is now accepted for `sens_init_constr` [r2100]
821  - fix typo in AMPL interface (by Weifeng Chen) [r2110]
822  - fix bug when compiling without AMPL interface [r2113]
823 - build system:
824  - updated instruction on using nowadays HSL sources (by T. Kelman)
825  - fixed issue with libdir being `<prefix>/lib64`
826 - other minor fixes
827 
828 ### 3.10.2 (2012-02-12)
829 
830 - updates to HSL interface (by Jonathan Hogg):
831  - MC68 can now be loaded dynamically, too
832  - MA86 exploits built-in scaling
833  - MA86 can choose best ordering from AMD and Metis
834  - fix for return code of MA86 for singular matrices
835 - corrected computation of Upsilon (norm of step SQUARED)
836 - updates to MSVS `v8-ifort` project files and addition of vc10
837  project files (using vc8-generated `IpoptFSS.dll`) (by Marcel Roelofs)
838 - minor bugfixes, include updates in BuildTools
839 
840 ### 3.10.1 (2011-09-20)
841 
842 - include updates in BuildTools, including new ThirdParty/Metis
843  (fix for URL to download Metis 4.0.3 release)
844 - linear solver loader prints error code when failing to load
845  library under Windows
846 - message on failure when reallocating memory in Mumps now includes
847  size of memory that was tried to be allocated
848 - added missing include of `cstdio/stdio.h` in `IpJournalist.hpp`
849 - minor fixes to build system
850 
851 ### 3.10.0 (2011-06-20)
852 
853 - move to new COIN-OR configuration and installation convention
854 - primal infeasibility output is now true infeasibility in original
855  problem formulation
856 
857 ## 3.9
858 
859 ### 3.9.3 (2011-04-07)
860 
861 - include updates in BuildTools, including new ThirdParty/Metis
862  (required to work with current metis release)
863 
864 ### 3.9.2 (2010-12-22)
865 
866 - converted from Common Public License to Eclipse Public License
867 - some bugfixes from BuildTools
868 
869 ### 3.9.1 (2010-11-26)
870 
871 - improved Hessian update for restoration phase
872 - added intermediate callback feature to C and Fortran interface
873 
874 ### 3.9.0 (2010-11-05)
875 
876 - switching to new BuildTools system
877 - added R interface (contributed by Jelmer Ypma)
878 - updates in AsNMPC (by Hans Pirnay)
879 
880 ## 3.8
881 
882 ### 3.8.3 (2010-06-29)
883 
884 - restated `SolveStatistics::TotalCPUTime` method for backward
885  compatibility
886 
887 ### 3.8.2 (2010-06-16)
888 
889 - uses MUMPS version 4.9 and Lapack version 3.2.1
890 - added AsNMPC contribution made by Hans Pirnay
891 - enhanced MA57 options
892 - several bug fixes and minor additions
893 
894 ### 3.8.1 (2009-10-30)
895 
896 - Bugfix in NLP function evaluation timing measurement. The
897  time for the objective function gradient had been forgotten.
898 
899 ### 3.8.0 (2009-10-30)
900 
901 - Added MSVC solution with Intel Fortran compiler to generate DLLs
902  (contributed by Marcel Roelofs). To make this work, a lot of methods
903  in exported headers have been made virtual
904 - changed default convergence tolerance in restoration phase (now same
905  as regular tolerance)
906 - output is flushed after each iteration
907 
908 ## 3.7
909 
910 ### 3.7.1 (2009-10-06)
911 
912 - bugfix for square problems
913 - correct timing information (obj gradient was forgotten)
914 - flush output buffer after each iteration
915 - first code for iterative WSMP version (experimental and undocumented)
916 
917 ### 3.7.0 (2009-07-16)
918 
919 - a number of fixes (including those from 2009 COIN-OR Bug Squashing Party)
920 - changes in some exposed header files to provide access to internal
921  data structures for specific applications
922 
923 ## 3.6
924 
925 ### 3.6.1 (2009-05-01)
926 
927 - minor corrections in tutorial files
928 
929 ### 3.6.0 (2009-04-29)
930 
931 - new Matlab interface
932 - added new option to limit cpu time: `max_cpu_time`
933 - added ThirdParty directory for Metis to be used with MUMPS or MA57
934 - updated CUTEr Makefile to make it work with CUTEr2
935 - added files for a tutorial (including coding exercise)
936 
937 ## 3.5
938 
939 ### 3.5.5 (2009-01-13)
940 
941 - minor fixes regarding compilation
942 - undocumented version of inexact method
943 
944 ### 3.5.4 (2008-09-29)
945 
946 - changed to MUMPS version 4.8.3 in externals (Mumps developers
947  seem to have removed 4.8.1).
948 
949 ### 3.5.3 (2008-09-19)
950 
951 - changed back to MUMPS version 4.8.1 since there seem to be issues
952  on Windows
953 
954 ### 3.5.2 (2008-09-18)
955 
956 - changed to latest version of MUMPS (4.8.2)
957 - some bugfixes (linear algebra objects, automatic problem scaling)
958 - made sure the subversion revision number is correct in all files
959 - allowed general additional data and cq in `IpData` and `IpCq`
960 
961 ### 3.5.1 (2008-08-26)
962 
963 - changed to latest version of MUMPS (4.8.1)
964 
965 ### 3.5.0 (2008-08-25)
966 
967 - added `ComputeRowAMax` and `ComputeColAMax` methods to Matrix base class
968 - changed externals for MUMPS to stable/1.1
969 - call `finalize_solution` in more failure cases
970  (this means that AMPL writes .sol file in more situations)
971 - added `IpTNLPReducer` as simple way to exclude constraints from problem
972 - several fixes, also from COIN-OR Bug Squashing Party 2008
973 
974 ## 3.4
975 
976 ### 3.4.2 (2008-07-18)
977 
978 - some bug fixes
979 - added wallclock time routine
980 - penalty function version does no longer crash if it
981  wants to go to restoration phase (not that this really helps
982  convergence though)
983 
984 ### 3.4.1 (2008-05-30)
985 
986 - some bug fixes
987 - deleted `v9` MSVC files again (since `v8` works fine for `v9`)
988 - print Ipopt version in default print level
989 - added option that allows to change name of options file
990  (`option_file_name`)
991 
992 ### 3.4.0 (2008-04-25)
993 
994 - added support to dynamically load HSL or Pardiso:
995  If Ipopt has been compiled without some HSL or Pardiso solver,
996  it can now load those solvers from a shared library at runtime
997  without recompilation. This will make distribution of binaries
998  easier. Does not work on all platforms yet.
999 - several bugfixes
1000 - ensured compilation of MSVS project files (`v8` and `v9`)
1001 - new special return code for square problems
1002  (`Feasible_Point_Found` returned if dual inf not small)
1003 - new initialization option for bound multipliers
1004  (see option `bound_mult_init_method`)
1005 - added simple penalty function line search option
1006  (`line_search_method=penalty`) - not guaranteed to converge, see
1007  Ipopt implementation paper (in MathProg)
1008 - some very basic method to approximate constraint Jacobian by
1009  finite differences (not efficient, but will hopefully be extended)
1010 
1011 ## 3.3
1012 
1013 ### 3.3.5 (2008-02-28)
1014 
1015 - corrected links for Ipopt mailing list
1016 - added missing `Makefile.in` for Matlab interface
1017 - the `addlibs*` files are now installed in `share/doc/coin/Ipopt`
1018  instead of lib
1019 - updates in Matlab interface
1020 - bugfix for ticket #56
1021 
1022 ### 3.3.4 (2007-12-27)
1023 
1024 - headers are now installed in `include/coin` (no longer in `include/ipopt`)
1025 - default for `dual_inf_tol` is now 1 (instead of 1e-4)
1026 - In matlab interface, here the text from Peter Carbonetto:
1027  There have been several significant changes made to the MATLAB interface
1028  since the last release. The most important two changes are:
1029  1. Following the "warm start" feature of IPOPT, you may pass in initial
1030  estimates for the Lagrange multipliers.
1031  2. Callback routines for computing the objective, gradient (etc.) are now
1032  specified using function handles instead of strings.
1033  (Thanks to Marcus Brubaker at the University of Toronto for the initial suggestion.)
1034 
1035 ### 3.3.3 (2007-09-25)
1036 
1037 - minor changes, bug fixes
1038 
1039 ### 3.3.1 (2007-06-20)
1040 
1041 Synchronized with all changes in trunk; probably more than to be
1042 remembered. In the following a few:
1043 - support for Mumps linear solver (contributed by Damian Hocking)
1044 - `--print-options` flag for ipopt ASL solver executable to see all
1045  Ipopt options (available through `ipopt.opt` file)
1046 - added Matlab interface (contributed by Peter Carbonetto)
1047 - added support for `f2c` compiler to compiler Fortran code with
1048  MSVC++ compiler
1049 - new MSVS support (now within MSVS project and also with `f2c`)
1050 - a number of small changes/bug fixes/improvements
1051 - small change in interface (e.g., `FinalizeSolution` method)
1052 
1053 ## 3.2
1054 
1055 ### 3.2.4 (2007-04-24)
1056 
1057 - updated download script for Blas to fit netlib's recent changes
1058 - using a more recent version of BuildTools
1059 
1060 ### 3.2.3 (2006-11-29)
1061 
1062 - updated download script for Lapack to fit to netlib's recent changes
1063 
1064 ### 3.2 r795 (2006-10-11)
1065 
1066 - Bugfix in L-BFGS update
1067 - fix in configure with detection of `sizeof(int*)` on Cygwin
1068 
1069 ### 3.2.1 (2006-07-14) - dev release number 764
1070 
1071 - Bugfix in least square multiplier estimate.
1072  It mainly showed up in LBFGS with restoration phase as seg fault
1073 
1074 ### 3.2.0 (2006-07-07) - dev release number 757
1075 
1076 - changed installation procedure and directory structure to
1077  conform with new COIN-OR convention
1078 
1079 ## 3.1
1080 
1081 ### 3.1.0 (2006-04-08) - dev release number 714
1082 
1083 Several bug-fixes, improvements and additions. In particular:
1084 - new quasi-Newton approximation using L-BFGS
1085 - interfaces to linear solver MA57, WSMP, Pardiso
1086  (MUMPS and TAUCS not yet completed)
1087 - derivative checker
1088 - unit test
1089 - configure supports compilation under Cygwin with native Windows compilers
1090 - ScalableExample
1091 - user call-back method in TNLP
1092 
1093 ## 3.0
1094 
1095 ### 3.0.1 (2005-12-04)
1096 
1097 - Several corrections to Windows files
1098 - Fix termination if number of iterations is exceeded in restoration phase
1099 
1100 ### 3.0.0 (2005-08-26) - dev release number 510
1101 
1102 - First official release of the new C++ implementation of Ipopt.
1103 
1104 ## 2.2
1105 
1106 ### no new release (2005-08-19)
1107 
1108 - corrected detection of BLAS libraries for SUN (make sure the example Makefiles work)
1109 - upgrade LICENSE file to CPL version 1.0 as retrieved from www.opensource.org
1110 
1111 ### 2.2.1e (2005-05-30)
1112 
1113 - fixed sign of multipliers returned to AMPL
1114  (bug reported by Rhoda Baker and Karsten Theissen)
1115 - switched to automake 1.9.5
1116 
1117 ### no new release (2005-01-07)
1118 
1119 - bugfix for the limited memory BFGS in case of square problems
1120  (bug reported by Wanhe Zhang and Ned Nedialkov)
1121 
1122 ### 2.2.1d (2004-10-05)
1123 
1124 - Added `outlev` as an option to the AMPL solver as a synonym for `iprint`
1125 - For `iprint` = 0, the output lines per iteration are now suppressed
1126 - corrected two bugs in `configure` script (test for size of `long` etc
1127  before Fortran libraries as added to `LIBS`; prevent cycling in
1128  `make -j 1` test)
1129 - internally renamed subroutine `ERROR` to `OPTERROR` (`ERROR` had a
1130  name clash for some Fortran compiler)
1131 - avoid uninitialized variable in `update_b_lm.f`
1132 - minor correction in computation of residual in `get_step_full.F`
1133 - minor change for slack correction in `filter.F`
1134 
1135 ### 2.2.1c (2004-07-20)
1136 
1137 - corrected bug leading to very small `QTAU` in rare circumstances
1138 
1139 ### 2.2.1b (2004-05-21)
1140 
1141 - Make `DFILLINFACT` option available through AMPL interface
1142 - Now, later increase of memory requirement for Harwell solvers is also
1143  based on `DFILLINFACT`, instead of using a fixed values of 10.
1144 
1145 ### 2.2.1a (2004-05-13)
1146 
1147 - fix in `IPOPT/ipopt/mainloop.F`:
1148  The multipliers were not scaled back for low printlevel.
1149 
1150 ### no new release (2004-04-28)
1151 
1152 - fix in `IPOPT/AMPL_interface/ipoptAMPL.c`:
1153  Now the mulitpliers for the constraints are passed back to AMPL.
1154 - Added download scripts using `wget` to get ASL, BLAS, and LAPACK more easily.
1155  Thanks to Frank Wuebbeling for the hint.
1156 
1157 ### 2.2.1 (2004-04-25)
1158 
1159 - AMPL solver executable is now called `ipopt` (instead of `ipoptAMPL`).
1160  This fixed also problem with assigning IPOPT options from within
1161  AMPL. (reported by Karsten Theissen)
1162 - default value for number of iterations is now 10000 (instead of 1000)
1163 - new option: `IMAXCPUSEC` to be set to the maximum number of CPU seconds
1164  after which the algorithm should stop. The check is performed only at
1165  certain points in the algorithm, so that the executable might run longer
1166  than specified.
1167  The algorithm also stops (as before) when the file "STOP" is detected in
1168  the current directory. Finally, a call to `USER_REQUESTED_STOP` has
1169  been added if the preprocessor macro `USE_USER_REQUESTED_STOP` has been
1170  defined. If this `LOGICAL` function returns `.TRUE.`, the algorithm also
1171  stops. This feature was requested by David Ternet.
1172 - if IPOPT is run several time in a row, the counting of function
1173  evaluations is restarted after every new call of IPOPT.
1174 - in `get_step_full`, now check if the number of negative eigenvalues is LESS
1175  than then number constraints. If so, increase the pivot tolerance,
1176  and if that doesn't help, pretend that the system is singular.
1177  (this fixed a problem reported by Hans Mittelmann)
1178 - decrease default values for `DPIVTOLMAX`
1179 - suppress superflous leading zeros in iteration output. For long output
1180  (`ioutput=1`) include CPU time
1181 - a few changes regarding the inertia correction (`get_step_full.F`).
1182  This decreases CPU time significantly in a few cases.
1183 - corrected problem in C-files related to names for `struct`'s (some GNU
1184  compiler complained)
1185 - changed default options for GNU compilers (now `-O3 -funroll-loops`,
1186  no longer `-O2` and `-mieee-fp`)
1187 - no reference to MC35 in `resto_tron` if `HAVE_MC35` is not defined
1188 - `#error` preprocessor directive removed from `*.F` files, since not all
1189  compilers understand it (reported by Hans Mittelmann)
1190 - switch to automake 1.8.3
1191 
1192 ### 2.2.0 (2004-03-10)
1193 
1194 Many things have changed since the last official release.
1195 Here a few highlights:
1196 - easier installation procedure with autoconf
1197 - algorithm made more robust and efficient
1198 - new restoration phase for filter method (TRON no longer needed for
1199  full-space option anymore)
1200 - C-interface