[feat] Do not use absolute paths in generated batch scripts

docs/advanced.rst

@@ -300,8 +300,8 @@ The generated job script for this example is the following:
    #SBATCH --job-name="prerun_demo_check_daint_gpu_PrgEnv-gnu"
    #SBATCH --time=0:10:0
    #SBATCH --ntasks=1
-   #SBATCH --output=/path/to/stage/gpu/prerun_demo_check/PrgEnv-gnu/prerun_demo_check.out
-   #SBATCH --error=/path/to/stage/gpu/prerun_demo_check/PrgEnv-gnu/prerun_demo_check.err
+   #SBATCH --output=prerun_demo_check.out
+   #SBATCH --error=prerun_demo_check.err
    #SBATCH --constraint=gpu
    module load daint-gpu
    module unload PrgEnv-cray

docs/tutorial.rst

@@ -296,8 +296,8 @@ The generated job script for the test case we are currently inspecting is the fo
   #SBATCH --job-name="rfm_Example1Test_job"
   #SBATCH --time=0:10:0
   #SBATCH --ntasks=1
-  #SBATCH --output=/path/to/stage/daint/gpu/PrgEnv-gnu/Example1Test/rfm_Example1Test_job.out
-  #SBATCH --error=/path/to/stage/daint/gpu/PrgEnv-gnu/Example1Test/rfm_Example1Test_job.err
+  #SBATCH --output=rfm_Example1Test_job.out
+  #SBATCH --error=rfm_Example1Test_job.err
   #SBATCH --constraint=gpu
   module load daint-gpu
   module unload PrgEnv-cray

reframe/core/schedulers/__init__.py

@@ -95,8 +95,8 @@ def __init__(self,
         self._num_cpus_per_task = num_cpus_per_task
         self._use_smt = use_smt
         self._script_filename = script_filename or '%s.sh' % name
-        self._stdout = stdout or os.path.join(workdir, '%s.out' % name)
-        self._stderr = stderr or os.path.join(workdir, '%s.err' % name)
+        self._stdout = stdout or '%s.out' % name
+        self._stderr = stderr or '%s.err' % name
         self._time_limit = time_limit
         self._nodelist = None
 

reframe/core/schedulers/pbs.py

@@ -130,7 +130,9 @@ def cancel(self):
 
     def finished(self):
         super().finished()
-        done = os.path.exists(self.stdout) and os.path.exists(self.stderr)
+        with os_ext.change_dir(self.workdir):
+            done = os.path.exists(self.stdout) and os.path.exists(self.stderr)
+
         if done:
             t_now = datetime.now()
             self._time_finished = self._time_finished or t_now

unittests/test_launchers.py

@@ -126,8 +126,8 @@ def expected_command(self):
     def expected_minimal_command(self):
         return ('srun '
                 '--job-name=fake_job '
-                '--output=./fake_job.out '
-                '--error=./fake_job.err '
+                '--output=fake_job.out '
+                '--error=fake_job.err '
                 '--ntasks=1 '
                 '--foo')
 

unittests/test_policies.py

@@ -183,8 +183,8 @@ def test_pass_in_retries(self):
 
 
 class TaskEventMonitor(executors.TaskEventListener):
-    """Event listener for monitoring the execution of the asynchronous execution
-    policy.
+    """Event listener for monitoring the execution of the asynchronous
+    execution policy.
 
     We need to make sure two things for the async policy:
 
@@ -240,9 +240,10 @@ def read_timestamps(self, tasks):
         self.begin_stamps = []
         self.end_stamps = []
         for t in tasks:
-            with open(evaluate(t.check.stdout), 'r') as f:
-                self.begin_stamps.append(float(f.readline().strip()))
-                self.end_stamps.append(float(f.readline().strip()))
+            with os_ext.change_dir(t.check.stagedir):
+                with open(evaluate(t.check.stdout), 'r') as f:
+                    self.begin_stamps.append(float(f.readline().strip()))
+                    self.end_stamps.append(float(f.readline().strip()))
 
         self.begin_stamps.sort()
         self.end_stamps.sort()