BF(TST): use caplog instead of capfd for testing if we log a warning

for some reason on my laptop capfd failed to provide desired effect. May be because
I am running   pytest with -s  ;)  will not investigate -- using caplog is the more
kosher way for this purpose, and allows for control of the logging level we would like
to have to ensure that the message is output

Author: yarikoptic

heudiconv/tests/test_convert.py

@@ -85,12 +85,14 @@ def test_update_uncombined_name():
     assert out_fn_test == out_fn_true
 
 
-def test_b0dwi_for_fmap(tmpdir, capfd):
+def test_b0dwi_for_fmap(tmpdir, caplog):
     """Make sure we raise a warning when .bvec and .bval files
     are present but the modality is not dwi.
     We check it by extracting a few DICOMs from a series with
     bvals: 5 5 1500
     """
+    import logging
+    caplog.set_level(logging.WARNING)
     tmppath = tmpdir.strpath
     subID = 'b0dwiForFmap'
     args = (
@@ -101,9 +103,8 @@ def test_b0dwi_for_fmap(tmpdir, capfd):
 
     # assert that it raised a warning that the fmap directory will contain
     # bvec and bval files.
-    output = capfd.readouterr().err.split('\n')
     expected_msg = DW_IMAGE_IN_FMAP_FOLDER_WARNING.format(folder=op.join(tmppath, 'sub-%s', 'fmap') % subID)
-    assert [o for o in output if expected_msg in o]
+    assert any(expected_msg in c.message for c in caplog.records)
 
     # check that both 'fmap' and 'dwi' directories have been extracted and they contain
     # *.bvec and a *.bval files