mx_graalpython_bisect.py: add WORKS criterion to bisect job

Author: cosminbasca

bisect-benchmark.ini

@@ -38,5 +38,6 @@ good = 1234cafebabe
 enterprise = true
 
 # Which result parameter should be used for comparisons, one of BEST, WORST,
-# AVG (all runs), AVG (no warmup)
+# AVG (all runs), AVG (no warmup), WORKS (bench failures are treated as
+# regressions)
 benchmark_criterion = BEST

mx.graalpython/mx_graalpython_bisect.py

@@ -238,7 +238,12 @@ def benchmark_callback(suite, commit):
         output = mx.OutputCapture()
         retcode = mx.run(shlex.split(args.benchmark_command), out=mx.TeeOutputCapture(output), nonZeroIsFatal=False)
         if retcode:
-            raise RuntimeError("Failed to execute benchmark for {}".format(commit))
+            if args.benchmark_criterion == 'WORKS':
+                return sys.maxsize
+            else:
+                raise RuntimeError("Failed to execute benchmark for {}".format(commit))
+        elif args.benchmark_criterion == 'WORKS':
+            return 0
         match = re.search(r'{}.*duration: ([\d.]+)'.format(re.escape(args.benchmark_criterion)), output.data)
         if not match:
             raise RuntimeError("Failed to get result from the benchmark")