Cumulus MFEM/MOOSE build scripts

scripts/cumulus_build_scripts/01_fetch_submodules_login-node.sh

@@ -0,0 +1,336 @@
+#!/bin/bash
+# =============================================================================
+# 01_fetch_submodules_loginnode.sh
+#
+# Pre-fetch all submodules and external package tarballs that the
+# MOOSE/MFEM build needs. Need to run on the login node (internet access)
+#
+# Hashes and URLs are read automatically from the checked-out repo,
+# so this script works with any MOOSE version.
+#
+# USAGE   : bash 01_fetch_submodules_loginnode.sh
+#           (from any directory — it finds MOOSE via $HOME/moose)
+# PS ./01_fetch_submodules_loginnode.sh won't work because it needs sudo rights
+# =============================================================================
+
+set -e
+set -u
+
+MOOSE_DIR="$HOME/moose"
+PKGS_DIR="$HOME/petsc_packages" #directory where 02 point at
+
+# Commands for better readable and pretty output formatting
+info() { echo "[INFO] $*"; }
+warn() { echo "[WARN] $*"; }
+error() { echo "[ERROR] $*"; exit 1; }
+step() {
+    echo ""
+    echo "=== $* ==="
+    echo "=================================================="
+}
+
+# prechecks
+step "Pre-flight checks"
+
+command -v git >/dev/null 2>&1 || error "git not found. Load it: module load git"
+
+[ -d "$MOOSE_DIR" ] || error "MOOSE not found at $MOOSE_DIR"
+[ -f "$MOOSE_DIR/.gitmodules" ] || error "Not a git repo: $MOOSE_DIR"
+
+if [ -n "${SLURM_JOB_ID:-}" ]; then
+    warn "SLURM_JOB_ID is set - you appear to be on a compute node."
+    warn "This script needs internet. Continue anyway? [y/N]"
+    read -r ans
+    [[ "$ans" =~ ^[Yy]$ ]] || exit 1
+fi
+
+info "Checking internet connectivity..."
+curl -s --max-time 5 https://github.com >/dev/null 2>&1 \
+    || error "Cannot reach github.com - run this on a login node."
+info "Internet OK."
+
+mkdir -p "$PKGS_DIR"
+
+# download resume and skips if already exists (function)
+download() {
+    local url="$1" out="$2"
+    if [ -f "$out" ] && [ -s "$out" ] && file "$out" | grep -qE "gzip|bzip2|XZ|Zip|tar"; then
+        info "Already present: $(basename "$out") ($(du -sh "$out" | cut -f1))"
+        return 0
+    fi
+    info "Downloading $(basename "$out") ..."
+    if command -v wget >/dev/null 2>&1; then
+        wget -q --show-progress -c "$url" -O "$out" \
+            || { warn "wget failed, retrying with curl..."; curl -L -o "$out" "$url"; }
+    else
+        curl -L -o "$out" "$url"
+    fi
+    if ! file "$out" | grep -qE "gzip|bzip2|XZ|Zip|tar"; then
+        rm -f "$out"
+        error "Download failed or returned non-tarball content for: $url"
+    fi
+    info "  OK: $(du -sh "$out" | cut -f1)"
+}
+
+# get hash of a submodule from git (function)
+submodule_hash() {
+    # Usage: submodule_hash <moose_dir> <submodule_path>
+    git -C "$1" ls-tree HEAD "$2" 2>/dev/null | awk '{print $3}'
+}
+
+# clone + checkout a submodule that has update=none function
+clone_pinned() {
+    local path="$1" url="$2" hash="$3" name="$4"
+    if [ -f "$path/CMakeLists.txt" ]; then
+        local current
+        current=$(git -C "$path" rev-parse HEAD 2>/dev/null || echo "unknown")
+        if [ "$current" = "$hash" ]; then
+            info "$name already at correct commit $hash - skipping."
+            return 0
+        else
+            warn "$name checked out at $current, expected $hash - re-cloning."
+        fi
+    fi
+    info "Cloning $name @ $hash ..."
+    rm -rf "$path"
+    git clone "$url" "$path"
+    cd "$path"
+    git fetch --all --quiet
+    git checkout "$hash"
+    git submodule update --init --recursive
+    info "$name cloned OK"
+}
+
+# SECTION 1 — PETSc submodule
+
+step "Section 1: PETSc submodule"
+
+cd "$MOOSE_DIR"
+git submodule update --init petsc
+info "PETSc submodule OK"
+
+# SECTION 2 — Determine PETSc external package URLs from configure.py
+# in the from the moose directory
+# Here PETSc configure in "offline dry-run" mode on the login node.
+# It exits with code 10 and prints the URLs it needs — then automagically
+# download them automatically.
+
+step "Section 2: PETSc external package tarballs (auto-detected URLs)"
+
+# Load python3 (may already be available or need a module)
+command -v python3 >/dev/null 2>&1 || error "python3 not found — load it: module load Python"
+
+info "Running PETSc configure in dry-run / download-dir mode to detect URLs..."
+info "This will exit with code 10 (normal) and print the list of packages."
+echo ""
+
+CONFIGURE_OUT=$(
+    cd "$MOOSE_DIR/petsc"
+    python3 ./configure \
+        --with-packages-download-dir="$PKGS_DIR" \
+        --with-debugging=no \
+        --with-mpi=1 \
+        --with-shared-libraries=1 \
+        --with-cxx-dialect=C++17 \
+        --with-fortran-bindings=0 \
+        --with-sowing=0 \
+        --download-fblaslapack=1 \
+        --download-hypre=1 \
+        --download-metis=1 \
+        --download-parmetis=1 \
+        --download-mumps=1 \
+        --download-scalapack=1 \
+        --download-superlu_dist=1 \
+        --download-slepc=1 \
+        --download-strumpack=1 \
+        --download-hdf5=1 \
+        --download-libceed=1 \
+        --download-openblas=1 \
+        --download-hpddm=1 \
+        --download-ptscotch=0 \
+        --download-kokkos=0 \
+        --download-kokkos-kernels=0 \
+        --download-umpire=0 \
+        2>&1 || true # exit 10 is expected
+)
+
+echo "$CONFIGURE_OUT" | grep -v "^Executing\|^stdout\|^$" | head -60 || true
+echo ""
+
+info "Parsing and downloading packages listed by PETSc configure..."
+
+# Parse URLs from PETSc output.
+# Output format (one package per line):
+#   pkgname ['git clone https://...', 'https://.../foo.tar.gz']
+#  extract only https:// tarball URLs (not git clone lines).
+while IFS= read -r line; do
+    # Match lines that start with a word then a space then ['
+    if echo "$line" | grep -qE "^[a-zA-Z_][a-zA-Z0-9_]* \['"; then
+        pkg=$(echo "$line" | awk '{print $1}')
+
+        # Pull out every https URL ending in .tar.gz or .tgz
+        # Use grep -oE (extended, no Perl) for broader compatibility
+        urls=$(echo "$line" | grep -oE "https://[^']+\.(tar\.gz|tgz)" || true)
+
+        if [ -z "$urls" ]; then
+            warn "No tarball URL found for $pkg — may need manual download"
+            continue
+        fi
+
+        # Prefer ANL mirror if available, otherwise take the first URL
+        url=$(echo "$urls" | grep "cels.anl.gov" | head -1)
+        [ -z "$url" ] && url=$(echo "$urls" | head -1)
+
+        # Safety: skip if URL is still empty or doesn't look like a URL
+        if [ -z "$url" ] || ! echo "$url" | grep -q "^https://"; then
+            warn "Skipping $pkg — could not extract a valid URL from: $line"
+            continue
+        fi
+
+        # Filename is the last path component of the URL
+        fname=$(basename "$url")
+
+        # Safety: skip if filenname is empty or looks like a directory
+        if [ -z "$fname" ] || [ -d "$PKGS_DIR/$fname" ]; then
+            warn "Skipping $pkg — could not determine output filename from URL: $url"
+            continue
+        fi
+
+        info "Package: $pkg  ->  $fname"
+        download "$url" "$PKGS_DIR/$fname"
+    fi
+done <<<"$CONFIGURE_OUT"
+
+info "All PETSc tarballs downloaded into $PKGS_DIR"
+echo ""
+ls -lh "$PKGS_DIR"
+
+# SECTION 3 — libMesh submodules (eigen, netgen)
+# netgen on cumulus module goes up to version GCCcore-12.3.0
+# Here version >13.0.0 is needed aparently
+step "Section 3: libMesh submodules"
+
+cd "$MOOSE_DIR"
+git submodule update --init --recursive libmesh
+info "libMesh submodules OK"
+
+# SECTION 4 — WASP
+step "Section 4: WASP submodule"
+
+cd "$MOOSE_DIR"
+git submodule update --init --recursive framework/contrib/wasp 2>/dev/null \
+    || info "WASP may be handled internally by update_and_rebuild_wasp.sh — continuing."
+
+# SECTION 5 — Conduit  (update = none in .gitmodules — must clone manually)
+step "Section 5: Conduit"
+
+CONDUIT_HASH=$(submodule_hash "$MOOSE_DIR" "framework/contrib/conduit")
+[ -n "$CONDUIT_HASH" ] || error "Could not read Conduit hash from git ls-tree."
+info "Pinned Conduit commit: $CONDUIT_HASH"
+
+# Resolve relative URL from .gitmodules  (../../LLNL/conduit.git)
+CONDUIT_URL="https://github.com/LLNL/conduit.git"
+
+clone_pinned \
+    "$MOOSE_DIR/framework/contrib/conduit" \
+    "$CONDUIT_URL" \
+    "$CONDUIT_HASH" \
+    "Conduit"
+
+# SECTION 6 — MFEM  (update = none in .gitmodules — must clone manually)
+step "Section 6: MFEM"
+
+MFEM_HASH=$(submodule_hash "$MOOSE_DIR" "framework/contrib/mfem")
+[ -n "$MFEM_HASH" ] || error "Could not read MFEM hash from git ls-tree."
+info "Pinned MFEM commit: $MFEM_HASH"
+MFEM_URL="https://github.com/mfem/mfem.git"
+
+clone_pinned \
+    "$MOOSE_DIR/framework/contrib/mfem" \
+    "$MFEM_URL" \
+    "$MFEM_HASH" \
+    "MFEM"
+
+# SECTION 7 — GSLIB  (MFEM fetches this at build time; pre-clone to avoid
+#              the compute-node network error during make)
+step "Section 7: GSLIB (required by MFEM)"
+
+# Read the tag/commit MFEM's CMakeLists expects
+MFEM_CMAKE="$MOOSE_DIR/framework/contrib/mfem/CMakeLists.txt"
+GSLIB_TAG=$(grep -A5 "gslib\|GSLIB" "$MFEM_CMAKE" 2>/dev/null \
+    | grep -iE "GIT_TAG|VERSION" | head -1 \
+    | grep -oE '[0-9]+\.[0-9]+\.[0-9]+' || echo "")
+
+GSLIB_DIR="$PKGS_DIR/gslib"
+GSLIB_URL="https://github.com/Nek5000/gslib.git"
+
+if [ -d "$GSLIB_DIR/.git" ]; then
+    info "GSLIB already cloned at $GSLIB_DIR"
+    cd "$GSLIB_DIR"
+    git fetch --all --quiet
+else
+    info "Cloning GSLIB into $GSLIB_DIR ..."
+    git clone "$GSLIB_URL" "$GSLIB_DIR"
+    cd "$GSLIB_DIR"
+fi
+
+if [ -n "$GSLIB_TAG" ]; then
+    info "Checking out GSLIB tag/version: $GSLIB_TAG"
+    git checkout "v${GSLIB_TAG}" 2>/dev/null \
+        || git checkout "$GSLIB_TAG" 2>/dev/null \
+        || warn "Could not checkout tag $GSLIB_TAG — using default branch"
+else
+    warn "Could not auto-detect GSLIB version from MFEM CMakeLists — using latest main"
+    git checkout main 2>/dev/null || git checkout master 2>/dev/null || true
+fi
+info "GSLIB OK  (stored at $GSLIB_DIR)"
+
+# SECTION 8 — HYPRE hash verification
+# (PETSc configure already handled the download above, but we double-check
+#  the exact commit hash that MFEM's FindHYPRE.cmake requires, if any)
+
+step "Section 8: Verify HYPRE hash"
+
+HYPRE_FILE=$(find "$PKGS_DIR" -maxdepth 1 -name "*hypre*" -type f 2>/dev/null | head -1)
+if [ -n "$HYPRE_FILE" ] && [ -s "$HYPRE_FILE" ]; then
+    info "HYPRE tarball present: $(basename "$HYPRE_FILE") ($(du -sh "$HYPRE_FILE" | cut -f1))"
+else
+    # PETSc configure should have got it; if not, grab it explicitly
+    warn "HYPRE tarball not detected — fetching manually..."
+    HYPRE_HASH=$(grep -r "hypre" "$MOOSE_DIR/petsc/config/BuildSystem/config/packages/HYPRE.py" \
+        2>/dev/null | grep -oE '[0-9a-f]{40}' | head -1 || echo "")
+    if [ -n "$HYPRE_HASH" ]; then
+        info "HYPRE hash from PETSc config: $HYPRE_HASH"
+        download \
+            "https://github.com/hypre-space/hypre/archive/${HYPRE_HASH}.tar.gz" \
+            "$PKGS_DIR/${HYPRE_HASH}.tar.gz"
+    else
+        warn "Could not auto-detect HYPRE hash — check $PKGS_DIR manually."
+    fi
+fi
+
+# FINAL SUMMARY
+# Print out summary with relevant info for 02_build_moose_computenode.sh
+# Also sanity check
+
+echo ""
+echo "============================================================"
+echo "  Pre-fetch complete!"
+echo "============================================================"
+echo ""
+echo "  Tarballs staging dir : $PKGS_DIR"
+echo "  Tarball count        : $(find "$PKGS_DIR" -maxdepth 1 -name '*.tar.gz' | wc -l)"
+du -sh "$PKGS_DIR"
+echo ""
+echo "  Submodule status:"
+git -C "$MOOSE_DIR" submodule status petsc libmesh \
+    framework/contrib/conduit framework/contrib/mfem 2>/dev/null || true
+echo ""
+echo "  GSLIB cached at      : $GSLIB_DIR"
+echo ""
+echo "  Next step:"
+echo "    Request a compute node:"
+echo "      srun --pty --time=08:00:00 --ntasks=8 --mem=32G bash"
+echo "    Then run:"
+echo "      bash 02_build_moose_computenode.sh"
+echo "============================================================"