Merge branch 'master' into better-tests

Author: sbryngelson

.claude/rules/fortran-conventions.md

@@ -23,13 +23,28 @@ Every Fortran module follows this pattern:
 
 ## Forbidden Patterns
 
-Caught by `./mfc.sh precheck` (source lint step 4/5):
+All checks below are enforced by `python3 toolchain/mfc/lint_source.py`
+(runs via `./mfc.sh precheck` and CI). See that file for the full list.
+
+Fortran/Fypp source (`src/`):
 - `dsqrt`, `dexp`, `dlog`, `dble`, `dabs`, `dcos`, `dsin`, `dtan`, etc. → use generic intrinsics
 - `1.0d0`, `2.5d-3` (Fortran `d` exponent literals) → use `1.0_wp`, `2.5e-3_wp`
 - `double precision` → use `real(wp)` or `real(stp)`
 - `real(8)`, `real(4)` → use `wp` or `stp` kind parameters
 - Raw `!$acc` or `!$omp` directives → use Fypp GPU_* macros from `parallel_macros.fpp`
-- Full list of forbidden patterns: `toolchain/bootstrap/precheck.sh`
+- `int(8._wp, ...)` hardcoded byte size → use `storage_size(0._stp)/8`
+- Bare integer kind like `2_wp` → use `2.0_wp`
+- Junk patterns (`...`, `---`, `===`) in code or comments (no separator comments)
+- Duplicate entries in Fypp `#:for ... in [...]` lists
+- Identical adjacent non-trivial lines (copy-paste bugs)
+
+Python (`examples/`, `benchmarks/`, `toolchain/`):
+- `===` separator comments → remove
+- `----` or longer separator comments → remove (3 dashes `---` is allowed for markdown)
+
+Shell (`.github/`, `toolchain/`):
+- `===` or `----` separator comments → remove
+- Echo separators longer than 20 characters → shorten
 
 Enforced by convention/code review (not automated):
 - `goto`, `COMMON` blocks, global `save` variables

.github/file-filter.yml

@@ -25,6 +25,7 @@ yml: &yml
   - '.github/workflows/phoenix/**'
   - '.github/workflows/frontier/**'
   - '.github/workflows/frontier_amd/**'
+  - '.github/workflows/common/**'
   - '.github/scripts/**'
   - '.github/workflows/bench.yml'
   - '.github/workflows/test.yml'
@@ -37,3 +38,6 @@ checkall: &checkall
   - *tests
   - *scripts
   - *yml
+
+cases_py:
+  - 'toolchain/mfc/test/cases.py'

.github/scripts/run_case_optimization.sh

@@ -36,9 +36,9 @@ for case in "${benchmarks[@]}"; do
     case_dir="$(dirname "$case")"
     case_name="$(basename "$case_dir")"
     echo ""
-    echo "========================================"
+    echo "===================="
     echo "Case-optimization test: $case_name"
-    echo "========================================"
+    echo "===================="
 
     # Clean any previous output
     rm -rf "$case_dir/D" "$case_dir/p_all" "$case_dir/restart_data"
@@ -65,11 +65,11 @@ for case in "${benchmarks[@]}"; do
 done
 
 echo ""
-echo "========================================"
+echo "===================="
 echo "Case-optimization summary: $passed passed, $failed failed"
 if [ $failed -gt 0 ]; then
     echo "Failed cases:$failed_cases"
 fi
-echo "========================================"
+echo "===================="
 
 [ $failed -eq 0 ] && exit 0 || exit 1

.github/scripts/run_parallel_benchmarks.sh

@@ -20,9 +20,9 @@ cluster="$3"
 # Get the directory where this script lives (pr/.github/scripts/)
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 
-echo "=========================================="
+echo "===================="
 echo "Starting benchmark jobs..."
-echo "=========================================="
+echo "===================="
 
 # For Phoenix GPU benchmarks, select a consistent GPU partition so PR and
 # master always land on the same GPU type.
@@ -82,9 +82,9 @@ if [ "${pr_exit}" -ne 0 ] || [ "${master_exit}" -ne 0 ]; then
     echo "WARNING: Benchmark jobs had failures: pr=${pr_exit}, master=${master_exit}"
     echo "Checking for partial results..."
 else
-    echo "=========================================="
+    echo "===================="
     echo "Both benchmark jobs completed successfully!"
-    echo "=========================================="
+    echo "===================="
 fi
 
 pr_yaml="pr/${job_slug}.yaml"

.github/scripts/submit-slurm-job.sh

@@ -185,6 +185,7 @@ job_device="$device"
 job_interface="$interface"
 job_shard="$shard"
 job_cluster="$cluster"
+export GITHUB_EVENT_NAME="$GITHUB_EVENT_NAME"
 
 . ./mfc.sh load -c $compiler_flag -m $module_mode
 

.github/scripts/submit_and_monitor_bench.sh

@@ -36,9 +36,9 @@ if [ ! -f "$yaml_file" ]; then
     echo ""
     output_file="${job_slug}.out"
     echo "[$dir] Last 100 lines of job output ($output_file):"
-    echo "----------------------------------------"
+    echo "--------------------"
     tail -n 100 "$output_file" 2>/dev/null || echo "  Could not read output file"
-    echo "----------------------------------------"
+    echo "--------------------"
     exit 1
 fi
 

.github/workflows/cleanliness.yml

@@ -86,6 +86,12 @@ jobs:
               grep -F 'Wmaybe-uninitialized' master.txt > mMaybe.txt
               diff prMaybe.txt mMaybe.txt
 
+          - name: Conversion Warnings Diff
+            continue-on-error: true
+            run: |
+              grep -F 'Wconversion' pr.txt > prConversion.txt
+              grep -F 'Wconversion' master.txt > mConversion.txt
+              diff prConversion.txt mConversion.txt
 
           - name: Everything Diff
             continue-on-error: true
@@ -106,12 +112,14 @@ jobs:
               pr_argument=$(grep -c -F 'Wunused-dummy-argument' pr.txt)
               pr_value=$(grep -c -F 'Wunused-value' pr.txt)
               pr_uninit=$(grep -c -F 'Wmaybe-uninitialized' pr.txt)
+              pr_conversion=$(grep -c -F 'Wconversion' pr.txt)
               pr_everything=$(grep -c '\-W' pr.txt)
 
               master_variable=$(grep -c -F 'Wunused-variable' master.txt)
               master_argument=$(grep -c -F 'Wunused-dummy-argument' master.txt)
               master_value=$(grep -c -F 'Wunused-value' master.txt)
               master_uninit=$(grep -c -F 'Wmaybe-uninitialized' master.txt)
+              master_conversion=$(grep -c -F 'Wconversion' master.txt)
               master_everything=$(grep -c '\-W' master.txt  )    
 
               echo "pr_everything=$pr_everything" >> $GITHUB_ENV
@@ -124,6 +132,7 @@ jobs:
               echo "Unused Dummy Argument: $pr_argument, Difference: $((pr_argument - master_argument))"
               echo "Unused Value: $pr_value, Difference: $((pr_value - master_value))"
               echo "Maybe Uninitialized: $pr_uninit, Difference: $((pr_uninit - master_uninit))"
+              echo "Conversion: $pr_conversion, Difference: $((pr_conversion - master_conversion))"
               echo "Everything: $pr_everything, Difference: $((pr_everything - master_everything))"
           
 

.github/workflows/common/rebuild-cache.sh

@@ -0,0 +1,23 @@
+#!/bin/bash
+set -e
+
+# Number of parallel jobs: use SLURM allocation or default to 24.
+# Cap at 64 to avoid overwhelming OpenMPI daemons and OS process limits with concurrent launches.
+NJOBS="${SLURM_CPUS_ON_NODE:-24}"
+if [ "$NJOBS" -gt 64 ]; then NJOBS=64; fi
+
+# Clean stale build artifacts: the self-hosted runner may have a cached
+# GPU build (e.g. --gpu mp) whose CMake flags are incompatible with gcov.
+./mfc.sh clean
+
+# Source retry_build() for NFS stale file handle resilience (3 attempts).
+source .github/scripts/retry-build.sh
+
+# Build MFC with gcov coverage instrumentation (CPU-only, gfortran).
+retry_build ./mfc.sh build --gcov -j 8
+
+# Run all tests in parallel, collecting per-test coverage data.
+# Each test gets an isolated GCOV_PREFIX directory so .gcda files
+# don't collide. Coverage is collected per-test after all tests finish.
+# --gcov is required so the internal build step preserves instrumentation.
+./mfc.sh test --build-coverage-cache --gcov -j "$NJOBS"

.github/workflows/common/test.sh

@@ -82,4 +82,10 @@ if [ -n "${job_shard:-}" ]; then
     shard_opts="--shard $job_shard"
 fi
 
-./mfc.sh test -v --max-attempts 3 -a -j $n_test_threads $rdma_opts $device_opts $build_opts $shard_opts -- -c $job_cluster
+# Only prune tests on PRs; master pushes must run the full suite.
+prune_flag=""
+if [ "${GITHUB_EVENT_NAME:-}" = "pull_request" ]; then
+    prune_flag="--only-changes"
+fi
+
+./mfc.sh test -v --max-attempts 3 $prune_flag -a -j $n_test_threads $rdma_opts $device_opts $build_opts $shard_opts -- -c $job_cluster