fix: set -e prevents catching vgpu-util failures

[andrewwhitecdw]

This PR addresses the following issue in rhel10/nvidia-driver: set -e prevents catching vgpu-util failures.

Changes

• rhel10/nvidia-driver: set -e prevents catching vgpu-util failures.

Details

--- a/rhel10/nvidia-driver
+++ b/rhel10/nvidia-driver
@@ -1,21 +1,19 @@
-    count=$(vgpu-util count)
-    if [ $? -ne 0 ]; then
-         echo "cannot find vgpu devices on host, pleae check /var/log/vgpu-util.log for more details..."
-         return 0
-    fi
-    NUM_VGPU_DEVICES=$(echo "$count" | awk -F= '{print $2}')
-    if [ $NUM_VGPU_DEVICES -eq 0 ]; then
-        # no vgpu devices found, treat as passthrough
-        return 0
-    fi
-    echo "found $NUM_VGPU_DEVICES vgpu devices on host"
-
-    # find compatible guest driver using driver catalog
-    if [ -d "/mnt/shared-nvidia-driver-toolkit/drivers" ]; then
-        drivers_path="/mnt/shared-nvidia-driver-toolkit/drivers"
-    fi
-    version=$(vgpu-util match -i "${drivers_path}" -c "${drivers_path}/vgpuDriverCatalog.yaml")
-    if [ $? -ne 0 ]; then
-        echo "cannot find match for compatible vgpu driver from available list, please check /var/log/vgpu-util.log for more details..."
-        return 1
-    fi
+    if ! count=$(vgpu-util count); then
+         echo "cannot find vgpu devices on host, pleae check /var/log/vgpu-util.log for more details..."
+         return 0
+    fi
+    NUM_VGPU_DEVICES=$(echo "$count" | awk -F= '{print $2}')
+    if [ $NUM_VGPU_DEVICES -eq 0 ]; then
+        # no vgpu devices found, treat as passthrough
+        return 0
+    fi
+    echo "found $NUM_VGPU_DEVICES vgpu devices on host"
+
+    # find compatible guest driver using driver catalog
+    if [ -d "/mnt/shared-nvidia-driver-toolkit/drivers" ]; then
+        drivers_path="/mnt/shared-nvidia-driver-toolkit/drivers"
+    fi
+    if ! version=$(vgpu-util match -i "${drivers_path}" -c "${drivers_path}/vgpuDriverCatalog.yaml"); then
+        echo "cannot find match for compatible vgpu driver from available list, please check /var/log/vgpu-util.log for more details..."
+        return 1
+    fi

Tests

• tests/test_sete_command_substitution.sh

--- /dev/null
+++ tests/test_sete_command_substitution.sh
@@ -0,0 +1,20 @@
+#!/bin/bash
+set -eu
+
+# Regression test: command substitution followed by a $? check is unreachable
+# under 'set -e'. The fixed code uses 'if ! cmd=$(...); then ... fi' instead.
+
+fail=0
+if grep -A1 -E '^[[:space:]]*kernel_module_type=\$\(' rhel10/nvidia-driver | grep -q '^[[:space:]]*if \[ \$\? -ne 0 \]'; then
+    echo "FAIL: nvidia-installer result is still checked with unreachable $?"
+    fail=1
+fi
+if grep -A1 -E '^[[:space:]]*(count|version)=\$\(' rhel10/nvidia-driver | grep -q '^[[:space:]]*if \[ \$\? -ne 0 \]'; then
+    echo "FAIL: vgpu-util result is still checked with unreachable $?"
+    fail=1
+fi
+
+if [ "$fail" -ne 0 ]; then
+    exit 1
+fi
+
+echo "PASS: command substitution results are checked correctly"

Contributor guidelines

Per this repo's CONTRIBUTING.md:

• All commits are signed off (Signed-off-by trailer, DCO).

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.