SpackCommand uses log_output to capture command output.

lib/spack/spack/main.py

@@ -34,9 +34,10 @@
 import inspect
 import pstats
 import argparse
-import tempfile
+from six import StringIO
 
 import llnl.util.tty as tty
+from llnl.util.tty.log import log_output
 from llnl.util.tty.color import *
 
 import spack
@@ -367,7 +368,7 @@ class SpackCommand(object):
         install('-v', 'mpich')
 
     Use this to invoke Spack commands directly from Python and check
-    their stdout and stderr.
+    their output.
     """
     def __init__(self, command):
         """Create a new SpackCommand that invokes ``command`` when called."""
@@ -376,9 +377,6 @@ def __init__(self, command):
         self.command_name = command
         self.command = spack.cmd.get_command(command)
 
-        self.returncode = None
-        self.error = None
-
     def __call__(self, *argv, **kwargs):
         """Invoke this SpackCommand.
 
@@ -389,24 +387,24 @@ def __call__(self, *argv, **kwargs):
             fail_on_error (optional bool): Don't raise an exception on error
 
         Returns:
-            (str, str): output and error as a strings
+            (str): combined output and error as a string
 
         On return, if ``fail_on_error`` is False, return value of comman
         is set in ``returncode`` property, and the error is set in the
         ``error`` property.  Otherwise, raise an error.
         """
+        # set these before every call to clear them out
+        self.returncode = None
+        self.error = None
+
         args, unknown = self.parser.parse_known_args(
             [self.command_name] + list(argv))
 
         fail_on_error = kwargs.get('fail_on_error', True)
 
-        out, err = sys.stdout, sys.stderr
-        ofd, ofn = tempfile.mkstemp()
-        efd, efn = tempfile.mkstemp()
-
+        out = StringIO()
         try:
-            sys.stdout = open(ofn, 'w')
-            sys.stderr = open(efn, 'w')
+            with log_output(out):
                 self.returncode = _invoke_spack_command(
                     self.command, self.parser, args, unknown)
 
@@ -418,25 +416,13 @@ def __call__(self, *argv, **kwargs):
             if fail_on_error:
                 raise
 
-        finally:
-            sys.stdout.flush()
-            sys.stdout.close()
-            sys.stderr.flush()
-            sys.stderr.close()
-            sys.stdout, sys.stderr = out, err
-
-            return_out = open(ofn).read()
-            return_err = open(efn).read()
-            os.unlink(ofn)
-            os.unlink(efn)
-
         if fail_on_error and self.returncode not in (None, 0):
             raise SpackCommandError(
                 "Command exited with code %d: %s(%s)" % (
                     self.returncode, self.command_name,
                     ', '.join("'%s'" % a for a in argv)))
 
-        return return_out, return_err
+        return out.getvalue()
 
 
 def _main(command, parser, args, unknown_args):

lib/spack/spack/test/cmd/dependencies.py

@@ -36,14 +36,14 @@
 
 
 def test_immediate_dependencies(builtin_mock):
-    out, err = dependencies('mpileaks')
+    out = dependencies('mpileaks')
     actual = set(re.split(r'\s+', out.strip()))
     expected = set(['callpath'] + mpis)
     assert expected == actual
 
 
 def test_transitive_dependencies(builtin_mock):
-    out, err = dependencies('--transitive', 'mpileaks')
+    out = dependencies('--transitive', 'mpileaks')
     actual = set(re.split(r'\s+', out.strip()))
     expected = set(
         ['callpath', 'dyninst', 'libdwarf', 'libelf'] + mpis + mpi_deps)
@@ -52,7 +52,7 @@ def test_transitive_dependencies(builtin_mock):
 
 def test_immediate_installed_dependencies(builtin_mock, database):
     with color_when(False):
-        out, err = dependencies('--installed', 'mpileaks^mpich')
+        out = dependencies('--installed', 'mpileaks^mpich')
 
     lines = [l for l in out.strip().split('\n') if not l.startswith('--')]
     hashes = set([re.split(r'\s+', l)[0] for l in lines])
@@ -65,7 +65,7 @@ def test_immediate_installed_dependencies(builtin_mock, database):
 
 def test_transitive_installed_dependencies(builtin_mock, database):
     with color_when(False):
-        out, err = dependencies('--installed', '--transitive', 'mpileaks^zmpi')
+        out = dependencies('--installed', '--transitive', 'mpileaks^zmpi')
 
     lines = [l for l in out.strip().split('\n') if not l.startswith('--')]
     hashes = set([re.split(r'\s+', l)[0] for l in lines])

lib/spack/spack/test/cmd/dependents.py

@@ -33,13 +33,13 @@
 
 
 def test_immediate_dependents(builtin_mock):
-    out, err = dependents('libelf')
+    out = dependents('libelf')
     actual = set(re.split(r'\s+', out.strip()))
     assert actual == set(['dyninst', 'libdwarf'])
 
 
 def test_transitive_dependents(builtin_mock):
-    out, err = dependents('--transitive', 'libelf')
+    out = dependents('--transitive', 'libelf')
     actual = set(re.split(r'\s+', out.strip()))
     assert actual == set(
         ['callpath', 'dyninst', 'libdwarf', 'mpileaks', 'multivalue_variant',
@@ -48,7 +48,7 @@ def test_transitive_dependents(builtin_mock):
 
 def test_immediate_installed_dependents(builtin_mock, database):
     with color_when(False):
-        out, err = dependents('--installed', 'libelf')
+        out = dependents('--installed', 'libelf')
 
     lines = [l for l in out.strip().split('\n') if not l.startswith('--')]
     hashes = set([re.split(r'\s+', l)[0] for l in lines])
@@ -64,7 +64,7 @@ def test_immediate_installed_dependents(builtin_mock, database):
 
 def test_transitive_installed_dependents(builtin_mock, database):
     with color_when(False):
-        out, err = dependents('--installed', '--transitive', 'fake')
+        out = dependents('--installed', '--transitive', 'fake')
 
     lines = [l for l in out.strip().split('\n') if not l.startswith('--')]
     hashes = set([re.split(r'\s+', l)[0] for l in lines])

lib/spack/spack/test/cmd/python.py

@@ -29,5 +29,5 @@
 
 
 def test_python():
-    out, err = python('-c', 'import spack; print(spack.spack_version)')
+    out = python('-c', 'import spack; print(spack.spack_version)')
     assert out.strip() == str(spack.spack_version)

lib/spack/spack/test/cmd/url.py

@@ -83,30 +83,30 @@ def test_url_with_no_version_fails():
 
 
 def test_url_list():
-    out, err = url('list')
+    out = url('list')
     total_urls = len(out.split('\n'))
 
     # The following two options should not change the number of URLs printed.
-    out, err = url('list', '--color', '--extrapolation')
+    out = url('list', '--color', '--extrapolation')
     colored_urls = len(out.split('\n'))
     assert colored_urls == total_urls
 
     # The following options should print fewer URLs than the default.
     # If they print the same number of URLs, something is horribly broken.
     # If they say we missed 0 URLs, something is probably broken too.
-    out, err = url('list', '--incorrect-name')
+    out = url('list', '--incorrect-name')
     incorrect_name_urls = len(out.split('\n'))
     assert 0 < incorrect_name_urls < total_urls
 
-    out, err = url('list', '--incorrect-version')
+    out = url('list', '--incorrect-version')
     incorrect_version_urls = len(out.split('\n'))
     assert 0 < incorrect_version_urls < total_urls
 
-    out, err = url('list', '--correct-name')
+    out = url('list', '--correct-name')
     correct_name_urls = len(out.split('\n'))
     assert 0 < correct_name_urls < total_urls
 
-    out, err = url('list', '--correct-version')
+    out = url('list', '--correct-version')
     correct_version_urls = len(out.split('\n'))
     assert 0 < correct_version_urls < total_urls
 
@@ -121,7 +121,7 @@ def test_url_summary():
     assert 0 < correct_versions <= sum(version_count_dict.values()) <= total_urls  # noqa
 
     # make sure it agrees with the actual command.
-    out, err = url('summary')
+    out = url('summary')
     out_total_urls = int(
         re.search(r'Total URLs found:\s*(\d+)', out).group(1))
     assert out_total_urls == total_urls