forked from pool/libminizinc
- update to 2.8.2:
* Bug fixes:
+ Fix incorrect FlatZinc definition in the gecode_presolver
solver library.
+ Fix type specialisation for function with an polymorphic
array argument and a non-array return type with the same
type-inst identifier.
+ Fix crash during typechecking of overloaded functions
differing only by inst returning structured types.
+ Fix incorrect type checking of polymorphic functions with
type-inst IDs inside structured types.
+ Fix evaluation error when boolean variables are fixed during
flattening.
+ Fix incorrect call names output when generating .ozn files.
+ Fix incorrect output of record access as tuple access in
.ozn files.
+ Fix definition of array_intersect so that it can be used in
non-positive contexts.
+ Fix standards definitions of increasing variants to correct
be ignored when the argument array is empty.
+ Fix a problem where exceptions thrown for undefined
expressions were not caught during the aggregation of linear
expressions, breaking relational semantics.
+ Fix crash when calculating computed domain of a declaration.
+ Fix x-y line/scatter visualisations.
* Changes:
+ Support var set``s of enums for ``array_union and
array_intersect.
+ Support var enums for the .. range operator.
+ Always perform substitution of fixed values before outputting
FlatZinc for consistency.
* Changes in the IDE:
+ Make process handling more robust to failures.
- update to 2.8.1:
* Bug fixes:
+ Fix incorrect message for par array out of bounds indicating
that array is empty
+ Fix incorrect propagation of Boolean variables potentially
causing output not accepted by Gecode and Chuffed.
+ Fix a problem where the usage of lb on a Boolean expression
would return -infinity..
+ Fix omission of error location when there is no stack trace
available.
+ Fix type specialisation to always make par versions of
functions available for output.
+ Fix internal error when checking return value of functions
involving arrays of optional values.
+ Fix incorrect false values for has_output_item when running
with --model-interface-only
+ Fix translation of search annotations with multi-dimensional
arrays as arguments.
+ Fix bug in output generation for some visualisation functions.
+ Fix problem where tuple or record assignments would sometimes
trigger segmentation faults.
+ Fix context when binding the result of flattening a
concatenation operation, resolving a possible segmentation
fault.
+ Fix incorrect possible evaluation error for in operator
involving an array RHS.
* Changes:
+ Add --solution-checker option to allow specifying the
solution checker (allowing use from a parameter configuration
.mpc file).
+ Produce tighter bounds in compute_mod_bounds for the
mod operator.
* Changes in the IDE:
+ Fix command used to run findMUS and Globalizer.
+ Add ability to set the ports used for the visualisation
server.
+ Add option for printing the visualisation server URL for
debugging purposes.
+ Add more information to subprocess error messages.
- update to 2.8.0:
* Changes in interfaces to solvers:
+ OR-Tools is now bundled with the MiniZinc IDE on all
platforms.
+ HiGHS is now loaded as a dynamic plugin instead of statically
linked, and now supports outputting intermediate solutions
for optimisation problems.
+ Add support for producing a JSON-based version of FlatZinc as
the output format of the MiniZinc compiler.
+ Replace supportsMzn, supportsFzn and supportsNL solver
configuration flags with new option inputType. The old flags
are still supported for backwards compatibility.
+ Add experimental support for restart based meta-search
specification for the Gecode and Chuffed solvers, as first
explored in Solver-Independent Large Neighbourhood Search and
A Modern Architecture for Constraint Modelling Languages.
+ Automatically detect current versions of CPLEX.
* Changes in the MiniZinc Library:
+ Add cumulatives scheduling global constraint.
+ Add the opt variants of the global_cardinality functions, and
add enumerated type typesafety between the values of the
decision variables and the cover parameters.
+ Add optional versions of count_* global constraints.
+ (strictly_)decreasing will now always be rewritten into
(strictly_)increasing and has support for option types.
+ Allow libraries to define how to iterate over var set of int
using the function set2iter, useful for different set
representations.
+ Stabilise the IDE visualisation library, allowing all
visualisation functions to be used from output statements,
and removing the need for manual calls to showJSON in custom
visualisations. This is a breaking change for users of the
previous experimental API.
+ Add mzn_half_reify_clause compiler option to allow solvers to
disable the half reification of bool_clause constraints.
+ Update the reflection functions ub, lb, and dom to return
enum values.
+ Use tuples to implement the decomposition of optional
variables, avoiding possible CSE aliasing problems.
* Changes in the compiler:
+ CSE has been adjusted to handle commutative functions when
annotated using the promise_commutative annotation.
+ mzn_reverse_map_var is now only called on output variables
that do not yet have reverse mappers after initial
flattening, but are required by the output model.
+ Improve error messaging for invalid parameter configuration
(.mpc) files.
+ Add a list of messages generated by solution checkers to the
checker message when using --json-stream.
+ Support output of command line argument parsing errors in
--json-stream mode.
* Bug fixes:
+ Fix restoration of tuple/record domains when flattening let
expressions.
+ Fix type checking error due to creation of incorrect par
versions of functions involving tuples/records.
+ Ensure that when --solver is used, the given solver
configuration file is always selected even when it collides
with a solver in the search paths.
+ Fix error when running satisfaction problems using the Gecode
presolver where an output variable is also an
introduced variable.
+ Resolve a problem where unification in the optimization phase
might remove variables part of the output.
+ Fix possible crash when printing the final solution using the
built-in Chuffed interface.
+ Don’t print the final status line from solution checkers.
+ Fix typechecking of par type-inst identifiers when
instantiated using structured types involving var types.
+ Implement fix and is_fixed for strucutred types.
+ Ensure reverse mappers are created when flattening
tuple/record literals. This resolves certain errors during
output processing in models using these types.
+ Fix problem where certain strings in exceptions might be
garbage collected before they are output.
+ Fix problem where argument names of generated functions could
conflict with model declared names.
+ Fix problem where the common type of complex records or
tuples was not correctly determined in the array literals.
+ Fix a problem in the parser where a nullptr would be used
before a syntax error was thrown.
+ Fix error management when reading preference files.
+ Fix segmentation fault caused by the creation of invalid
domain constraints generated for functions with arrays of
tuples or records parameters.
+ Fix crash when instantiating a type-inst identifier using
only <>.
+ Fix evaluation of comprehensions containing opt string.
+ Fix crash when instantiating polymorphic functions using
tuples/records with many var or opt fields.
+ Do not generate default DZN output when there are only
sectioned output items present.
+ Fix the edge indexing of the edges for the final node in the
neural_net global constraint.
+ Add better error messaging when an index set mismatch is
encountered in the standard library.
+ Fix evaluation error when indexing into an array using <>.
+ Fix incorrect unsatisfiability when equating optional
variables with non-intersecting domains.
+ Fix array access using optional indices for multidimensional
arrays.
+ Fix output of zero length arrays of optional variables.
+ Fix output processing performance degradation when printing
arrays in DZN output mode.
+ Fix card function on set of floats and bools.
+ Make set literals containing strings or annotations a
type error.
* Changes in the IDE:
+ Fix unreadable cheat sheet font colour in dark mode.
+ Add option to output objective value and enable by default.
+ Show manually input parameters in output window.
+ Fix missing checker messages.
+ Fix incorrect OpenSSL version in Linux packages.
- Add minimal checks
- Add patches:
* 0001-Disambiguate-UNKNOWN-EXCEPTION.patch
* 0002-Catch-std-exception-by-const-reference.patch
* 0003-Try-to-get-some-more-information-when-catching.patch
OBS-URL: https://build.opensuse.org/request/show/1135808
OBS-URL: https://build.opensuse.org/package/show/science/libminizinc?expand=0&rev=9
34 lines
1.2 KiB
Diff
34 lines
1.2 KiB
Diff
From c1e1cc1f76cac901752a870ac44a2ebc45729d96 Mon Sep 17 00:00:00 2001
|
|
From: =?UTF-8?q?Stefan=20Br=C3=BCns?= <stefan.bruens@rwth-aachen.de>
|
|
Date: Tue, 19 Jul 2022 19:31:23 +0200
|
|
Subject: [PATCH 2/3] Catch std::exception by const-reference
|
|
|
|
Although the details are ignored and do not matter, catching
|
|
std::exception by value is in general bad due to potential slicing,
|
|
and the potentially required copy.
|
|
---
|
|
include/minizinc/flatten.hh | 4 ++--
|
|
1 file changed, 2 insertions(+), 2 deletions(-)
|
|
|
|
diff --git a/include/minizinc/flatten.hh b/include/minizinc/flatten.hh
|
|
index 2b631e82..b51c9625 100644
|
|
--- a/include/minizinc/flatten.hh
|
|
+++ b/include/minizinc/flatten.hh
|
|
@@ -87,11 +87,11 @@ struct FlatteningOptions {
|
|
try {
|
|
std::random_device rdev("/dev/urandom");
|
|
seeds.push_back(rdev());
|
|
- } catch (std::exception) {
|
|
+ } catch (const std::exception&) {
|
|
try {
|
|
std::random_device rdev;
|
|
seeds.push_back(rdev());
|
|
- } catch (std::exception) { /* NOLINT(bugprone-empty-catch) */
|
|
+ } catch (const std::exception&) { /* NOLINT(bugprone-empty-catch) */
|
|
}
|
|
}
|
|
auto highrestime = static_cast<long unsigned int>(
|
|
--
|
|
2.37.0
|
|
|