Fix benchmark: detect failed commands and install cairo deps

- measure() now verifies command exit code; reports FAIL instead of
  a misleadingly fast time when a command crashes
- avg_runs() handles individual failures gracefully
- Install libcairo2-dev on Ubuntu for cairosvg PNG generation
- Fix applies to both CI workflow and local run.sh

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: brunoborges

.github/workflows/benchmark.yml

@@ -28,6 +28,10 @@ jobs:
         with:
           python-version: '3.13'
 
+      - name: Install system dependencies (Ubuntu)
+        if: runner.os == 'Linux'
+        run: sudo apt-get update && sudo apt-get install -y libcairo2-dev
+
       - name: Install Python dependencies
         run: pip install pyyaml cairosvg
 
@@ -47,19 +51,38 @@ jobs:
           # Timing helper using bash TIMEFORMAT
           measure() {
             TIMEFORMAT='%R'
-            local t
-            t=$( { time "$@" > /dev/null 2>&1; } 2>&1 )
+            local t ec
+            t=$( { time "$@" > /dev/null 2>&1; ec=$?; } 2>&1 )
+            if [[ ${ec:-1} -ne 0 ]]; then
+              # Re-check: the subshell exit code isn't always captured
+              if ! "$@" > /dev/null 2>&1; then
+                echo "FAIL"
+                return 1
+              fi
+            fi
             echo "$t"
           }
 
           avg_runs() {
             local n="$1"; shift
-            local sum=0
+            local sum=0 failures=0
             for ((i = 1; i <= n; i++)); do
               local t
-              t=$(measure "$@")
+              t=$(measure "$@") || true
+              if [[ "$t" == "FAIL" ]]; then
+                ((failures++)) || true
+                continue
+              fi
               sum=$(awk "BEGIN {print $sum + $t}")
             done
+            local success=$((n - failures))
+            if [[ $success -eq 0 ]]; then
+              echo "FAIL"
+              return 0
+            fi
+            awk "BEGIN {printf \"%.2f\", $sum / $success}"
+          }
+            done
             awk "BEGIN {printf \"%.2f\", $sum / $n}"
           }
 
@@ -194,6 +217,10 @@ jobs:
         with:
           python-version: '3.13'
 
+      - name: Install system dependencies (Ubuntu)
+        if: runner.os == 'Linux'
+        run: sudo apt-get update && sudo apt-get install -y libcairo2-dev
+
       - name: Install Python dependencies
         run: pip install pyyaml cairosvg
 
@@ -212,8 +239,14 @@ jobs:
 
           measure() {
             TIMEFORMAT='%R'
-            local t
-            t=$( { time "$@" > /dev/null 2>&1; } 2>&1 )
+            local t ec
+            t=$( { time "$@" > /dev/null 2>&1; ec=$?; } 2>&1 )
+            if [[ ${ec:-1} -ne 0 ]]; then
+              if ! "$@" > /dev/null 2>&1; then
+                echo "FAIL"
+                return 1
+              fi
+            fi
             echo "$t"
           }
 

html-generators/benchmark/run.sh

@@ -30,29 +30,35 @@ UPDATE_MD=false
 # ---------------------------------------------------------------------------
 # Helpers
 # ---------------------------------------------------------------------------
-time_once() {
-  /usr/bin/time -p "$@" > /dev/null 2>&1 | awk '/^real/ {print $2}'
-  # fallback: capture from stderr
-  local t
-  t=$( { /usr/bin/time -p "$@" > /dev/null; } 2>&1 | awk '/^real/ {print $2}' )
-  echo "$t"
-}
-
 measure() {
   local t
   t=$( { /usr/bin/time -p "$@" > /dev/null; } 2>&1 | awk '/^real/ {print $2}' )
+  # Verify the command actually succeeded
+  if ! "$@" > /dev/null 2>&1; then
+    echo "FAIL"
+    return 1
+  fi
   echo "$t"
 }
 
 avg_runs() {
   local n="$1"; shift
-  local sum=0
+  local sum=0 failures=0
   for ((i = 1; i <= n; i++)); do
     local t
-    t=$(measure "$@")
+    t=$(measure "$@") || true
+    if [[ "$t" == "FAIL" ]]; then
+      ((failures++)) || true
+      continue
+    fi
     sum=$(echo "$sum + $t" | bc)
   done
-  echo "scale=2; $sum / $n" | bc | sed 's/^\./0./'
+  local success=$((n - failures))
+  if [[ $success -eq 0 ]]; then
+    echo "FAIL"
+    return 0
+  fi
+  echo "scale=2; $sum / $success" | bc | sed 's/^\./0./'
 }
 
 # ---------------------------------------------------------------------------