diff options
Diffstat (limited to 'numpy/f2py/tests/test_crackfortran.py')
| -rw-r--r-- | numpy/f2py/tests/test_crackfortran.py | 78 |
1 files changed, 73 insertions, 5 deletions
diff --git a/numpy/f2py/tests/test_crackfortran.py b/numpy/f2py/tests/test_crackfortran.py index dcf8760db..49bfc13af 100644 --- a/numpy/f2py/tests/test_crackfortran.py +++ b/numpy/f2py/tests/test_crackfortran.py @@ -1,7 +1,10 @@ +import importlib import codecs +import time +import unicodedata import pytest import numpy as np -from numpy.f2py.crackfortran import markinnerspaces +from numpy.f2py.crackfortran import markinnerspaces, nameargspattern from . import util from numpy.f2py import crackfortran import textwrap @@ -132,6 +135,7 @@ class TestMarkinnerspaces: assert markinnerspaces("a 'b c' 'd e'") == "a 'b@_@c' 'd@_@e'" assert markinnerspaces(r'a "b c" "d e"') == r'a "b@_@c" "d@_@e"' + class TestDimSpec(util.F2PyTest): """This test suite tests various expressions that are used as dimension specifications. @@ -241,6 +245,7 @@ class TestModuleDeclaration: assert len(mod) == 1 assert mod[0]["vars"]["abar"]["="] == "bar('abar')" + class TestEval(util.F2PyTest): def test_eval_scalar(self): eval_scalar = crackfortran._eval_scalar @@ -257,13 +262,76 @@ class TestFortranReader(util.F2PyTest): def test_input_encoding(self, tmp_path, encoding): # gh-635 f_path = tmp_path / f"input_with_{encoding}_encoding.f90" - # explicit BOM is required for UTF8 - bom = {'utf-8': codecs.BOM_UTF8}.get(encoding, b'') with f_path.open('w', encoding=encoding) as ff: - ff.write(bom.decode(encoding) + - """ + ff.write(""" subroutine foo() end subroutine foo """) mod = crackfortran.crackfortran([str(f_path)]) assert mod[0]['name'] == 'foo' + + +class TestUnicodeComment(util.F2PyTest): + sources = [util.getpath("tests", "src", "crackfortran", "unicode_comment.f90")] + + @pytest.mark.skipif( + (importlib.util.find_spec("charset_normalizer") is None), + reason="test requires charset_normalizer which is not installed", + ) + def test_encoding_comment(self): + self.module.foo(3) + + +class TestNameArgsPatternBacktracking: + @pytest.mark.parametrize( + ['adversary'], + [ + ('@)@bind@(@',), + ('@)@bind @(@',), + ('@)@bind foo bar baz@(@',) + ] + ) + def test_nameargspattern_backtracking(self, adversary): + '''address ReDOS vulnerability: + https://github.com/numpy/numpy/issues/23338''' + trials_per_batch = 12 + batches_per_regex = 4 + start_reps, end_reps = 15, 25 + for ii in range(start_reps, end_reps): + repeated_adversary = adversary * ii + # test times in small batches. + # this gives us more chances to catch a bad regex + # while still catching it before too long if it is bad + for _ in range(batches_per_regex): + times = [] + for _ in range(trials_per_batch): + t0 = time.perf_counter() + mtch = nameargspattern.search(repeated_adversary) + times.append(time.perf_counter() - t0) + # our pattern should be much faster than 0.2s per search + # it's unlikely that a bad regex will pass even on fast CPUs + assert np.median(times) < 0.2 + assert not mtch + # if the adversary is capped with @)@, it becomes acceptable + # according to the old version of the regex. + # that should still be true. + good_version_of_adversary = repeated_adversary + '@)@' + assert nameargspattern.search(good_version_of_adversary) + + +class TestFunctionReturn(util.F2PyTest): + sources = [util.getpath("tests", "src", "crackfortran", "gh23598.f90")] + + def test_function_rettype(self): + # gh-23598 + assert self.module.intproduct(3, 4) == 12 + + +class TestFortranGroupCounters(util.F2PyTest): + def test_end_if_comment(self): + # gh-23533 + fpath = util.getpath("tests", "src", "crackfortran", "gh23533.f") + try: + crackfortran.crackfortran([str(fpath)]) + except Exception as exc: + assert False, f"'crackfortran.crackfortran' raised an exception {exc}" |