handle both darwin and gnu time

Author: Acribbs

cgatcore/pipeline/execution.py

@@ -27,6 +27,7 @@
 import math
 import shutil
 import gevent
+import shlex
 import cgatcore.experiment as E
 import cgatcore.iotools as iotools
 
@@ -43,6 +44,9 @@
     HAS_DRMAA = False
 import platform
 
+# Define TIME_CMD based on gtime availability
+TIME_CMD = shutil.which("gtime") or "time"
+
 # global drmaa session
 GLOBAL_SESSION = None
 
@@ -960,20 +964,16 @@ def _convert(key, v):
         return JobInfo(jobId=process.pid, resourceUsage=data)
 
     def run(self, statement_list):
-
         benchmark_data = []
         for statement in statement_list:
             self.logger.info("running statement:\n%s" % statement)
 
             full_statement, job_path = self.build_job_script(statement)
 
-            time_command = "gtime" if platform.system() == "Darwin" else "time"
-
-            # max_vmem is set to max_rss, not available by /usr/bin/time
+            # Use `shlex.quote()` to wrap `job_path` safely in the full statement
             full_statement = (
-                f"\\{time_command} --output={job_path}.times "
-                f"-f '"
-                f"exit_status\t%x\n"
+                f"{shlex.quote(TIME_CMD)} --output={shlex.quote(job_path + '.times')} "
+                f"-f 'exit_status\t%x\n"
                 f"user_t\t%U\n"
                 f"sys_t\t%S\n"
                 f"wall_t\t%e\n"
@@ -995,57 +995,42 @@ def run(self, statement_list):
                 f"socket_sent\t%s\n"
                 f"major_page_fault\t%F\n"
                 f"unshared_data\t%D\n' "
-                f"{job_path}"
+                f"{shlex.quote(job_path)}"
             )
 
-            while 1:
-                start_time = time.time()
-
-                os.environ.update(
-                    {'BASH_ENV': os.path.join(os.environ['HOME'], '.bashrc')})
-                process = subprocess.Popen(
-                    full_statement,
-                    cwd=self.work_dir,
-                    shell=True,
-                    stdin=subprocess.PIPE,
-                    stdout=subprocess.PIPE,
-                    stderr=subprocess.PIPE,
-                    env=os.environ.copy(),
-                    close_fds=True,
-                    executable="/bin/bash")
-
-                # process.stdin.close()
-                stdout, stderr = process.communicate()
-
-                end_time = time.time()
-
-                if process.returncode == 126:
-                    self.logger.warn(
-                        "repeating execution: message={}".format(stderr))
-                    time.sleep(1)
-                    continue
-
-                break
-            stdout = stdout.decode("utf-8")
-            stderr = stderr.decode("utf-8")
+            start_time = time.time()
+            os.environ.update({'BASH_ENV': os.path.join(os.environ['HOME'], '.bashrc')})
+            process = subprocess.Popen(
+                full_statement,
+                cwd=self.work_dir,
+                shell=True,
+                stdin=subprocess.PIPE,
+                stdout=subprocess.PIPE,
+                stderr=subprocess.PIPE,
+                env=os.environ.copy(),
+                close_fds=True,
+                executable="/bin/bash"
+            )
+
+            stdout, stderr = process.communicate()
+            end_time = time.time()
 
             if process.returncode != 0 and not self.ignore_errors:
                 raise OSError(
                     "---------------------------------------\n"
                     "Child was terminated by signal %i: \n"
                     "The stderr was: \n%s\n%s\n"
                     "-----------------------------------------" %
-                    (-process.returncode, stderr, statement))
+                    (-process.returncode, stderr.decode("utf-8"), statement)
+                )
 
-            resource_usage = self.collect_metric_data(process,
-                                                      start_time,
-                                                      end_time,
-                                                      time_data_file=job_path + ".times")
+            resource_usage = self.collect_metric_data(
+                process, start_time, end_time, time_data_file=job_path + ".times"
+            )
 
             benchmark_data.extend(
-                self.collect_benchmark_data(
-                    [statement],
-                    resource_usage=[resource_usage]))
+                self.collect_benchmark_data([statement], resource_usage=[resource_usage])
+            )
 
             try:
                 os.unlink(job_path)