[bugfix] Fix remote auto-detection crash

reframe/frontend/autodetect.py

@@ -8,6 +8,7 @@
 import os
 import shutil
 import tempfile
+import traceback
 
 import reframe as rfm
 import reframe.core.runtime as runtime
@@ -19,6 +20,10 @@
 from reframe.utility.cpuinfo import cpuinfo
 
 
+# This is meant to be used by the unit tests
+_TREAT_WARNINGS_AS_ERRORS = False
+
+
 def _contents(filename):
     '''Return the contents of a file.'''
 
@@ -36,7 +41,6 @@ def _log_contents(filename):
 class _copy_reframe:
     def __init__(self, prefix):
         self._prefix = prefix
-        self._prefix = runtime().get_option('general/0/remote_workdir')
         self._workdir = None
 
     def __enter__(self):
@@ -82,6 +86,9 @@ def _load_info(filename, schema=None):
         with open(filename) as fp:
             return _validate_info(json.load(fp), schema)
     except OSError as e:
+        if _TREAT_WARNINGS_AS_ERRORS:
+            raise
+
         getlogger().warning(
             f'could not load file: {filename!r}: {e}'
         )
@@ -101,9 +108,14 @@ def _save_info(filename, topo_info):
         with open(filename, 'w') as fp:
             json.dump(topo_info, fp, indent=2)
     except OSError as e:
+        if _TREAT_WARNINGS_AS_ERRORS:
+            raise
+
         getlogger().warning(
             f'could not save topology file: {filename!r}: {e}'
         )
+    else:
+        getlogger().debug(f'> saved topology in {filename!r}')
 
 
 def _is_part_local(part):
@@ -143,7 +155,11 @@ def _emit_script(job, env):
                 _log_contents(job.stderr)
                 topo_info = json.loads(_contents('topo.json'))
     except Exception as e:
+        if _TREAT_WARNINGS_AS_ERRORS:
+            raise
+
         getlogger().warning(f'failed to retrieve remote processor info: {e}')
+        getlogger().debug(traceback.format_exc())
 
     return topo_info
 
@@ -226,14 +242,11 @@ def detect_topology():
 
                 _save_info(topo_file, part.processor.info)
             elif detect_remote_systems:
-                with runtime.temp_environment(modules=temp_modules,
-                                              variables=temp_vars):
+                with runtime.temp_environment(modules=modules, variables=vars):
                     part._processor = ProcessorInfo(_remote_detect(part))
 
                 if part.processor.info:
                     _save_info(topo_file, part.processor.info)
 
-            getlogger().debug(f'> saved topology in {topo_file!r}')
-
         if not found_devinfo:
             getlogger().debug(f'> device auto-detection is not supported')

unittests/test_autodetect.py

@@ -7,16 +7,17 @@
 import os
 import pytest
 
-
+import reframe.frontend.autodetect as autodetect
+import unittests.utility as test_util
 from reframe.core.runtime import runtime
-from reframe.frontend.autodetect import detect_topology
 from reframe.utility.cpuinfo import cpuinfo
 
 
 @pytest.fixture
-def exec_ctx(make_exec_ctx_g, tmp_path, monkeypatch):
+def temp_topo(tmp_path, monkeypatch):
     # Monkey-patch HOME, since topology is always written there
     monkeypatch.setenv('HOME', str(tmp_path))
+    monkeypatch.setattr(autodetect, '_TREAT_WARNINGS_AS_ERRORS', True)
 
     # Create a devices file manually, since it is not auto-generated
     meta_prefix = tmp_path / '.reframe' / 'topology' / 'generic-default'
@@ -30,13 +31,12 @@ def exec_ctx(make_exec_ctx_g, tmp_path, monkeypatch):
             }
         ], fp)
 
-    yield from make_exec_ctx_g()
-
 
 @pytest.fixture
-def invalid_topo_exec_ctx(make_exec_ctx_g, tmp_path, monkeypatch):
+def invalid_topo(tmp_path, monkeypatch):
     # Monkey-patch HOME, since topology is always written there
     monkeypatch.setenv('HOME', str(tmp_path))
+    monkeypatch.setattr(autodetect, '_TREAT_WARNINGS_AS_ERRORS', True)
 
     # Create invalid processor and devices files
     meta_prefix = tmp_path / '.reframe' / 'topology' / 'generic-default'
@@ -47,11 +47,30 @@ def invalid_topo_exec_ctx(make_exec_ctx_g, tmp_path, monkeypatch):
     with open(meta_prefix / 'devices.json', 'w') as fp:
         fp.write('{')
 
+
+@pytest.fixture
+def default_exec_ctx(make_exec_ctx_g, temp_topo):
+    yield from make_exec_ctx_g()
+
+
+@pytest.fixture
+def remote_exec_ctx(make_exec_ctx, temp_topo):
+    if test_util.USER_CONFIG_FILE is None:
+        pytest.skip('no user configuration file supplied')
+
+    ctx = make_exec_ctx(test_util.USER_CONFIG_FILE,
+                        test_util.USER_SYSTEM,
+                        {'general/remote_detect': True})
+    yield ctx
+
+
+@pytest.fixture
+def invalid_topo_exec_ctx(make_exec_ctx_g, invalid_topo):
     yield from make_exec_ctx_g()
 
 
-def test_autotect(exec_ctx):
-    detect_topology()
+def test_autotect(default_exec_ctx):
+    autodetect.detect_topology()
     part = runtime().system.partitions[0]
     assert part.processor.info == cpuinfo()
     if part.processor.info:
@@ -80,7 +99,18 @@ def test_autotect(exec_ctx):
 
 
 def test_autotect_with_invalid_files(invalid_topo_exec_ctx):
-    detect_topology()
+    autodetect.detect_topology()
     part = runtime().system.partitions[0]
     assert part.processor.info == cpuinfo()
     assert part.devices == []
+
+
+def test_remote_autodetect(remote_exec_ctx):
+    # All we can do with this test is to trigger the remote auto-detection
+    # path; since we don't know what the remote user system is, we cannot test
+    # if the topology is right.
+    partition = test_util.partition_by_scheduler()
+    if not partition:
+        pytest.skip('job submission not supported')
+
+    autodetect.detect_topology()