Meh, the first version contained out-of-date patches.
these patches correct diagnostics dealing with variables in specification
expressions.
The first patch is a testsuite change, which fixes invalid specification
expressions that the second patch would diagnose.
The second patch removes a spurious diagnostic when a dummy procedure is
involved, and enables more valid ones, as visible in the testcases from the
first patch.
The patch is not completely trivial, and fix what is not really a regression,
so it is more for stage1, I think.
Tested for regression on x86_64-pc-linux-gnu. Ok for master when stage1
opens?
Mikael
v1 -> v2 changes:
- Fix condition guarding sym->result access.
Mikael Morin (2):
testsuite: Declare fortran array bound variables
fortran: Fix specification expression error with dummy procedures
[PR111781]
gcc/fortran/expr.cc | 8 +-
gcc/fortran/gfortran.h | 4 +-
gcc/fortran/resolve.cc | 77 +++++++++----------
gcc/fortran/symbol.cc | 58 ++++++++++++++
.../gfortran.dg/graphite/pr107865.f90 | 2 +-
gcc/testsuite/gfortran.dg/pr101267.f90 | 2 +-
gcc/testsuite/gfortran.dg/pr112404.f90 | 2 +-
gcc/testsuite/gfortran.dg/pr78061.f | 2 +-
gcc/testsuite/gfortran.dg/pr79315.f90 | 6 +-
gcc/testsuite/gfortran.dg/spec_expr_8.f90 | 24 ++++++
gcc/testsuite/gfortran.dg/spec_expr_9.f90 | 19 +++++
gcc/testsuite/gfortran.dg/vect/pr90681.f | 2 +-
gcc/testsuite/gfortran.dg/vect/pr97761.f90 | 2 +-
gcc/testsuite/gfortran.dg/vect/pr99746.f90 | 2 +-
14 files changed, 152 insertions(+), 58 deletions(-)
create mode 100644 gcc/testsuite/gfortran.dg/spec_expr_8.f90
create mode 100644 gcc/testsuite/gfortran.dg/spec_expr_9.f90
--
2.43.0
