Improve handling of nested and dummy leading comments, now properly ignores comments inside the docstring
Implement rudimentary pragmas for linter
Ensure the docstring cursor is always a PETSc linter Cursor, and improve note on no defined paramteres
Handle in-out and same-name function parameters correctly
Don't print extra newline at the end of notes
Allow [](ch_whatever) constructs to not be enclosed by backticks
Reformat the error message output to remove the cursor blurb from the start, it wasconfusing and misleading, and add support for color diagnostics
Support in-out parameter descriptions
Need to have test harness remove -options_left to tests that test handling of errors in testsCommit-type: testing
When the user does not properly handle a returned error code and another error is detected, print a very useful message indicating what has happended.Reported-by: Stefano Zampini
Merge branch 'barry/2023-06-15/add-fortran-file-viewer' into 'main'Allow one to attach a PETSCVIEWERASCII to a Fortran file unitSee merge request petsc/petsc!6607
petscviewer.h90 is missing from Fortran modules, add to petscsysmod.
Fix failing linter test
Merge remote-tracking branch 'origin/release'
Fix make check when --prefix points to a read-only file systemIn general make check should not be writing anything in a directory indicated by $PETSC_DIR since that is where PETSc is installed.Sa
Fix make check when --prefix points to a read-only file systemIn general make check should not be writing anything in a directory indicated by $PETSC_DIR since that is where PETSc is installed.Satish, we should have a make install that then uses chmod u-w newpetscdir ; make check to verify this bug never reappearsReported-by: Drew Parsons
show more ...
Add PetscCheck() and PetscCheckA() for FortranAlso fix incorrect use of " in Fortran examples that should be '
test(sys): Add sys/tests/ex17.cTests PetscSynchronized*()
Manpages: collective info should go after synopsis
SYS tests: fixes for FPE
Implement the following optimizations in Petsc::UnorderedMap:1. Use std::optional to hold the values if C++17 is available. Erased elements are not fully destroyed in the implementation (they ar
Implement the following optimizations in Petsc::UnorderedMap:1. Use std::optional to hold the values if C++17 is available. Erased elements are not fully destroyed in the implementation (they are only marked as destroyed and then default re-contructed). Using std::optional allows us to skip calling the possibly non-trivial destructors for these "destroyed" elements on e.g. resize.2. Add shrink_to_fit() which prunes the capacity of the map to size(). Judicious use of this function, combined with the above change, makes resize() and further modification even more efficient.
Fix incorrect syntax in make check with mpi4py
Fix some documentation and typos
12345678910>>...13