Mentions légales du service

Skip to content
_Note:_ this release requires the version [v20.24](https://gitlab.inria.fr/MoReFEM/ThirdPartyCompilationFactory/-/tags/v20.24) of third party libraries: a new test about a PETSc wrapper fail with PETSc 3.12 (with a message the functionality is missing for a given type of matrix) but works as expected with 3.13.

## Model - two step parallelism

- Feature #1443: Two step parallelism now works - but currently you still need to prepare the data in parallel.
- Bug #1522: Mesh constructor from prepartitioned data was poorly implemented and was not working as expected for higher order.
- Feature #1531: Adapt FromVertexMatching operator to two step parallelism

## Operators

- Feature #1520 Implement a utility operator to write out local quadrature points into global coordinates.

## Geometry

- Design #1528: Go back to using shared_ptr for Coords.
- Bug #1526: The geometric elements related to boundary conditions were not properly kept during the reduction process in a parallel run.

## Core

- Support #1524: Remove most of the mandatory fields in input data.

## Utilities

- Support #1537: Add a create_if_necessary behaviour for FilesystemNS::Directory, and clarify ignore one.
- Support #1538: Use optional instead of unique_ptr for behaviour in FilesystemNS::Directory constructor, which makes a much leaner interface.
- Support #1539: Add tests for functions defined in Utilities::String - and modify slightly ConvertCharArray() to make it pass the test (one line was undue...).
- Support #1540 Add a Split() utility function for strings.

## ThirdParty

- Support #1532 Petsc wrapper: Refactor slightly the possible viewers to add more safety and enable reloading a matrix from a file
- Feature #1533: Add a PETSc wrapper around MatEqual.
- __Design #1535__: For PETSc Matrix And Vector, Internal() is no longer const.
- __Design #1536__: Remove MatrixOperations that were half baked and not used (to my knowledge).

## PostProcessing

- Bug #1519: The case of an Unknown present in several NumberingSubset was not properly foreseen.

## CI

- Support #1523: Replace the Alpine Docker image by a Fedora one.
- Support #1272 - #1534  Reactivate tests that were manual ones and were never updated after the test toochain was introduced (respectively about input data and PETSc matrix operations).
- Support #1525: Apply #1524 to PETSc VectorIO test: it no longer uses up a Model under the hood - instead a much lighter ToyModel class is used - and this class inherits from nothing.
- Support #1542: If a branch name includes 'nobuild', the build_and_test and check_warnings stages are skipped.
- Support #1541: Some modifications to make the analysis tools work properly (Valgrind test will nonetheless be in orange: the new version
of one of the library (probably Openmpi) gets a tiny memory leak (8o...) that I can't suppress due to a lack of information.

## ModelInstances

- Support #1530: Reduce the number of time iterations in hyperelastic model.

## Miscellaneous

- Bug #1527 In configure_cmake.py script, use the intended third party libraries (previously it was always the one chosen for release mode that was used).

- Support #1528: Modifications related to the upgrade of most of the third-party librairies (tag v20.24](https://gitlab.inria.fr/MoReFEM/ThirdPartyCompilationFactory/-/tags/v20.24) of [ThirdPartyCompilationFctory](https://gitlab.inria.fr/MoReFEM/ThirdPartyCompilationFactory) ).