Clean up GNU compiler warnings

CLAUDE.md

@@ -0,0 +1,234 @@
+# CLAUDE.md
+Claude Code Operating Rules for This Repository
+
+This file defines **hard constraints** and **preferred behaviors** for Claude
+working in this repository. These rules exist to protect numerical correctness,
+performance, and long-term maintainability of a high-order spectral element
+codebase.
+
+Follow all rules below unless explicitly instructed otherwise by a human
+maintainer.
+
+---
+
+## 1. Project Purpose
+
+This project implements **high-order spectral element methods** for solving
+systems of conservation laws (e.g., Euler, shallow water, Maxwell, MHD) on
+structured and unstructured meshes.
+
+Primary goals:
+- Numerical correctness and stability
+- Conservation properties
+- High performance on CPUs and GPUs
+- MPI scalability
+
+Changes must NOT degrade:
+- Accuracy order
+- Stability properties
+- Parallel scaling
+- Memory behavior
+
+---
+
+## 2. Fortran Standards & Toolchain
+
+### Language Standard
+- Target: **Fortran 2008**
+- Code must remain compatible with:
+  - gfortran ≥ 11
+  - ifx
+  - nvfortran
+  - amdflang
+
+### Prohibited Language Features
+Do NOT introduce:
+- Coarrays
+- Fortran 2018+ features
+- Compiler-specific extensions
+- Automatic polymorphism in performance-critical paths
+
+### Formatting & Conventions
+- Free-form source
+- `implicit none` required in all program units
+- Explicit `intent(in|out|inout)` on all dummy arguments
+- Lowercase keywords preferred
+- Line length ≤ 100 characters
+
+---
+
+## 3. Numerical & Algorithmic Constraints
+
+### Hard Rules
+- Do NOT change the mathematical formulation without approval
+- Do NOT change discretization order
+- Do NOT change basis, quadrature, or nodal ordering
+- Do NOT reorder floating-point reductions
+- Do NOT alter time integration schemes
+
+### Floating-Point Behavior
+- Bitwise reproducibility may be required
+- Preserve operation ordering in loops
+- Avoid algebraic “simplifications” unless mathematically justified
+
+### Array Semantics
+- Do NOT replace explicit loops with array syntax unless equivalence is proven
+- Avoid implicit temporaries
+
+---
+
+## 4. Performance Rules (Critical)
+
+This is an HPC codebase. Performance regressions are unacceptable.
+
+### Memory
+- Avoid temporary allocations in hot paths
+- No automatic arrays in tight loops
+- No hidden allocations via array slicing
+
+### Loops
+- Preserve loop ordering for cache locality
+- Do NOT replace DO loops with WHERE / FORALL
+- Vectorization-friendly structure must be preserved
+
+### Abstraction
+- Do NOT introduce runtime polymorphism in kernels
+- Avoid excessive modularization inside hot loops
+
+---
+
+## 5. Parallel Programming
+
+### MPI
+- MPI calls must remain explicit
+- Do NOT introduce blocking collectives inside time-stepping loops
+- Do NOT change communicator usage
+- Preserve rank-local data ownership
+
+### OpenMP / GPU
+- Preserve OpenMP semantics
+- Do NOT move data regions without explicit instruction
+- GPU kernels must preserve memory access patterns
+- No implicit host/device transfers
+
+---
+
+## 6. Code Organization & APIs
+
+### File & Module Structure
+- Do NOT rename modules
+- Do NOT move files between directories
+- Do NOT change public interfaces without approval
+- Preserve module dependency order
+
+### Public APIs
+- Public procedures are considered **stable**
+- Backward compatibility is required unless stated otherwise
+
+---
+
+## 7. Testing & Validation
+
+### Required
+- All existing regression tests must pass
+- Do NOT modify reference output files
+- Numerical differences must be justified
+
+### New Code
+- New features require:
+  - A test case
+  - Clear validation criteria
+- MPI tests must work on ≥ 2 ranks
+
+---
+
+## 8. Documentation & Comments
+
+### Preserve Scientific Meaning
+- Do NOT remove comments describing:
+  - Equations
+  - Algorithms
+  - Numerical assumptions
+
+### New Routines
+Must include:
+- Mathematical description
+- Variable meaning and units
+- Expected input ranges
+
+---
+
+## 9. Code Formatting
+
+All Fortran source must be formatted with [`fprettify`](https://pypi.org/project/fprettify/) using the project's `fprettify.config`. PRs are checked for formatting before any other tests run.
+
+To format all source files manually:
+
+```shell
+fprettify './src/'      --config-file ./fprettify.config --recursive --case 1 1 1 1
+fprettify './test/'     --config-file ./fprettify.config --recursive --case 1 1 1 1
+fprettify './examples/' --config-file ./fprettify.config --recursive --case 1 1 1 1
+```
+
+Alternatively, install the provided `pre-commit` hook to apply formatting automatically on each commit:
+
+```shell
+pip install pre-commit fprettify
+pre-commit install   # run from repository root
+```
+
+When editing Fortran files, apply `fprettify` before committing. Do NOT manually reformat code by hand in ways that deviate from `fprettify` output.
+
+---
+
+## 10. Prohibited Actions (Explicit)
+
+Do NOT:
+- Rewrite code in another language
+- Convert procedural code to OO Fortran
+- Replace MPI with coarrays
+- Introduce external dependencies
+- “Modernize” syntax without benefit
+- Delete legacy code without explanation
+
+---
+
+## 11. Domain-Specific Assumptions
+
+- Grid indexing follows project conventions (do NOT reorder indices)
+- Jacobians and metric terms are precomputed
+- Flux routines assume nodal basis ordering
+- Element-local operations must remain element-local
+- Halo exchange patterns are fixed
+
+---
+
+## 12. Preferred Behavior
+
+DO:
+- Ask before changing algorithms
+- Explain numerical and performance implications
+- Provide minimal diffs
+- Reference existing patterns in the codebase
+- Flag any uncertainty explicitly
+
+DO NOT:
+- Make large refactors unless requested
+- Assume intent beyond the explicit request
+
+---
+
+## 13. When in Doubt
+
+If a change could affect:
+- Numerical accuracy
+- Stability
+- Performance
+- Parallel behavior
+
+STOP and ask for clarification.
+
+---
+
+End of CLAUDE.md
+

examples/linear_euler2d_planewave_propagation.f90

@@ -107,6 +107,7 @@ pure function hbc2d_Prescribed_lineareuler2d_planewave(this,x,t) result(exts)
     exts(2) = u*shape ! u
     exts(3) = v*shape ! v
     exts(4) = p*shape ! pressure
+    if(.false.) exts(1) = exts(1)+t ! suppress unused-dummy-argument warning
 
   endfunction hbc2d_Prescribed_lineareuler2d_planewave
 

examples/linear_euler2d_planewave_reflection.f90

@@ -122,6 +122,7 @@ pure function hbc2d_Prescribed_lineareuler2d_planewave(this,x,t) result(exts)
     exts(2) = u*(shi-shr) ! u
     exts(3) = v*(shi+shr) ! v
     exts(4) = p*(shi+shr) ! pressure
+    if(.false.) exts(1) = exts(1)+t ! suppress unused-dummy-argument warning
 
   endfunction hbc2d_Prescribed_lineareuler2d_planewave
 

examples/linear_shallow_water2d_kelvinwaves.f90

@@ -38,13 +38,11 @@ program linear_shallow_water2d_kelvinwaves
   real(prec),parameter :: f0 = 10.0_prec ! reference coriolis parameter (1/s)
   real(prec),parameter :: Cd = 0.25_prec ! Linear drag coefficient (1/s)
   real(prec),parameter :: iointerval = 0.05 ! Write files 20 times per characteristic time scale
-  real(prec) :: r
   real(prec) :: e0,ef ! Initial and final entropy
   type(LinearShallowWater2D) :: modelobj ! Shallow water model
   type(Lagrange),target :: interp ! Interpolant
   type(Mesh2D),target :: mesh ! Mesh class
   type(SEMQuad),target :: geometry ! Geometry class
-  integer :: i,j,iel
   real(prec),parameter :: g = 1.0_prec ! Acceleration due to gravity
   real(prec),parameter :: H = 1.0_prec ! Uniform resting depth
   character(LEN=255) :: WORKSPACE

examples/linear_shallow_water2d_nonormalflow.f90

@@ -43,7 +43,6 @@ program linear_shallow_water2d_nonormalflow_model
   integer :: bcids(1:4) ! Boundary conditions for structured mesh
   type(Mesh2D),target :: mesh ! Mesh class
   type(SEMQuad),target :: geometry ! Geometry class
-  character(LEN=255) :: WORKSPACE ! Used for file I/O
 
   real(prec),parameter :: g = 1.0_prec ! Acceleration due to gravity
   real(prec),parameter :: H = 1.0_prec ! Uniform resting depth

src/SELF_DGModel1D_t.f90

@@ -299,7 +299,7 @@ subroutine CalculateEntropy_DGModel1D_t(this)
     implicit none
     class(DGModel1D_t),intent(inout) :: this
     ! Local
-    integer :: iel,i,ivar
+    integer :: iel,i
     real(prec) :: e,s(1:this%solution%nvar),J
 
     e = 0.0_prec
@@ -323,7 +323,6 @@ subroutine setboundarycondition_DGModel1D_t(this)
     implicit none
     class(DGModel1D_t),intent(inout) :: this
     ! local
-    integer :: ivar
     integer :: N,nelem
     real(prec) :: x
 
@@ -627,10 +626,6 @@ subroutine Read_DGModel1D_t(this,fileName)
     character(*),intent(in) :: fileName
     ! Local
     integer(HID_T) :: fileId
-    integer(HID_T) :: solOffset(1:3)
-    integer :: firstElem
-    integer :: N
-
     call Open_HDF5(fileName,H5F_ACC_RDWR_F,fileId)
     call ReadArray_HDF5(fileId,'/controlgrid/solution/interior',this%solution%interior)
     call Close_HDF5(fileId)
@@ -643,14 +638,12 @@ subroutine WriteTecplot_DGModel1D_t(this,filename)
     class(DGModel1D_t),intent(inout) :: this
     character(*),intent(in),optional :: filename
     ! Local
-    character(8) :: zoneID
     integer :: fUnit
     integer :: iEl,i,iVar
     character(LEN=self_FileNameLength) :: tecFile
     character(LEN=self_TecplotHeaderLength) :: tecHeader
     character(LEN=self_FormatLength) :: fmat
     character(13) :: timeStampString
-    character(5) :: rankString
     type(Scalar1D) :: solution
     type(Scalar1D) :: x
     type(Lagrange),target :: interp

src/SELF_DGModel2D_t.f90

@@ -92,10 +92,6 @@ subroutine Init_DGModel2D_t(this,mesh,geometry)
     type(Mesh2D),intent(in),target :: mesh
     type(SEMQuad),intent(in),target :: geometry
     ! Local
-    integer :: ivar
-    character(LEN=3) :: ivarChar
-    character(LEN=25) :: varname
-
     this%mesh => mesh
     this%geometry => geometry
     call this%SetNumberOfVariables()
@@ -650,7 +646,7 @@ subroutine Read_DGModel2D_t(this,fileName)
     integer(HID_T) :: fileId
     integer(HID_T) :: solOffset(1:3)
     integer :: firstElem
-    integer :: N,ivar
+    integer :: ivar
 
     if(this%mesh%decomp%mpiEnabled) then
       call Open_HDF5(fileName,H5F_ACC_RDWR_F,fileId, &

src/SELF_DGModel3D_t.f90

@@ -92,10 +92,6 @@ subroutine Init_DGModel3D_t(this,mesh,geometry)
     type(Mesh3D),intent(in),target :: mesh
     type(SEMHex),intent(in),target :: geometry
     ! Local
-    integer :: ivar
-    character(LEN=3) :: ivarChar
-    character(LEN=25) :: varname
-
     this%mesh => mesh
     this%geometry => geometry
     call this%SetNumberOfVariables()

src/SELF_Geometry_2D.f90

@@ -263,7 +263,7 @@ subroutine WriteTecplot_SEMQuad(this,filename)
     ! Local
     character(8) :: zoneID
     integer :: fUnit
-    integer :: iEl,i,j,iVar
+    integer :: iEl,i,j
     character(LEN=self_TecplotHeaderLength) :: tecHeader
     character(LEN=self_FormatLength) :: fmat
 

src/SELF_Geometry_3D.f90

@@ -412,7 +412,7 @@ subroutine WriteTecplot_SEMHex(this,filename)
     ! Local
     character(8) :: zoneID
     integer :: fUnit
-    integer :: iEl,i,j,k,iVar
+    integer :: iEl,i,j,k
     character(LEN=self_TecplotHeaderLength) :: tecHeader
     character(LEN=self_FormatLength) :: fmat
 

src/SELF_LinearEuler2D_t.f90

@@ -126,8 +126,6 @@ pure function hbc2d_NoNormalFlow_LinearEuler2D_t(this,s,nhat) result(exts)
     real(prec),intent(in) :: s(1:this%nvar)
     real(prec),intent(in) :: nhat(1:2)
     real(prec) :: exts(1:this%nvar)
-    ! Local
-    integer :: ivar
 
     exts(1) = s(1) ! density
     exts(2) = (nhat(2)**2-nhat(1)**2)*s(2)-2.0_prec*nhat(1)*nhat(2)*s(3) ! u
@@ -150,6 +148,7 @@ pure function flux2d_LinearEuler2D_t(this,s,dsdx) result(flux)
     flux(3,2) = s(4)/this%rho0 ! y-velocity, y flux; p/rho0
     flux(4,1) = this%c*this%c*this%rho0*s(2) ! pressure, x flux : rho0*c^2*u
     flux(4,2) = this%c*this%c*this%rho0*s(3) ! pressure, y flux : rho0*c^2*v
+    if(.false.) flux(1,1) = flux(1,1)+dsdx(1,1) ! suppress unused-dummy-argument warning
 
   endfunction flux2d_LinearEuler2D_t
 
@@ -186,6 +185,7 @@ pure function riemannflux2d_LinearEuler2D_t(this,sL,sR,dsdx,nhat) result(flux)
     fR(4) = rho0*c*c*(u*nhat(1)+v*nhat(2)) ! pressure
 
     flux(1:4) = 0.5_prec*(fL(1:4)+fR(1:4))+c*(sL(1:4)-sR(1:4))
+    if(.false.) flux(1) = flux(1)+dsdx(1,1) ! suppress unused-dummy-argument warning
 
   endfunction riemannflux2d_LinearEuler2D_t
 
@@ -207,7 +207,7 @@ subroutine SphericalSoundWave_LinearEuler2D_t(this,rhoprime,Lr,x0,y0)
     real(prec),intent(in) ::  rhoprime,Lr,x0,y0
     ! Local
     integer :: i,j,iEl
-    real(prec) :: x,y,rho,r,E
+    real(prec) :: x,y,rho,r
 
     print*,__FILE__," : Configuring weak blast wave initial condition. "
     print*,__FILE__," : rhoprime = ",rhoprime