subversion-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Johan Corveleyn <jcor...@gmail.com>
Subject Re: Test failures with Python 3 (Re: PMCs: any Hackathon requests? (deadline 11 October))
Date Tue, 15 Oct 2019 23:07:32 GMT
On Tue, Oct 15, 2019 at 6:20 PM Yasuhito FUTATSUKI <futatuki@poem.co.jp> wrote:
>
> On 2019-10-15 17:17, Johan Corveleyn wrote:
> > On Tue, Oct 15, 2019 at 9:26 AM Yasuhito FUTATSUKI <futatuki@poem.co.jp> wrote:
> >>
> >> On 2019-10-15 07:04, Daniel Shahaf wrote:
> >>> Yasuhito FUTATSUKI wrote on Sun, 13 Oct 2019 04:01 +00:00:
> >>>> On 2019/10/13 7:24, Daniel Shahaf wrote:
> >>
> >>>> I see. Now I agree it would suffice here.
> >>>>
> >>>>> So, how about:
> >>>>>
> >>>>> 1. Make the test use non-binary mode for changing and reading the
> >>>>>       file 'lambda'.
>
> >>>>> 2. Locally revert the C part of r1841731 and make sure the modified
test
> >>>>>       still (correctly) fails.  (That revision both added the test
and
> >>>>>       fixed the bug the test checks for.)
>
> I overlooked comments in this test. On step 2 the test will continue to loop
> as far as resource is available, or until signaled.
>
> And yes, after 'svn merge -r1841731:1841730 subversion/libsvn_client/conflicts.c',
> the test can't reach patched line. So no more test is needed on step 2, both
> on Unix/Linux and on Windows.
>
> >>>>
> >>>> So it looks sufficient to me.
> >>>
> >>> Cool.  Will you perchance have time to do this?  No worries if not.
> >>
> >> Yes, I'll do it on FreeBSD on tonight or tomorrow night (in JST :)).
>
> with the attached patch, both with Python 2.7.15 and Python 3.7.0 on FreeBSD,
> tree_conflict_tests passed.
>
> >> However I think it is also need to test for each 1 and 2 on Windows,
> >> because r1841736 and r1841743 also were attempt to fix this test
> >> on Windows, with Python 2.
> >
> > Feel free to let me know if I need to test something on Windows.
>
> Thank you. Could you please test the tree_conflict_tests with this patch,
> both with Python 2 and Python 3 on Windows?

Okay, I can confirm that tree_conflict_tests works with Python 2.7.16
(both with and without the patch) and with 3.7.5 (with the patch) on
Windows 7.

I did have some trouble testing it with Python 3.7 though:

- First, I had to try it on the swig-py3 branch, because on trunk I
get this when trying to run any test with py 3.7:
[[[
Traceback (most recent call last):
  File "win-tests.py", line 134, in <module>
    cp.items('options'))
  File "build\generator\gen_win_dependencies.py", line 306, in __init__
    self.find_libraries(False)
  File "build\generator\gen_win_dependencies.py", line 327, in find_libraries
    self._find_jdk(show_warnings)
  File "build\generator\gen_win_dependencies.py", line 1085, in _find_jdk
    vermatch = re.search(r'(([0-9]+(\.[0-9]+)+)(_[._0-9]+)?)', line, re.M)
  File "C:\Python37\lib\re.py", line 183, in search
    return _compile(pattern, flags).search(string)
TypeError: cannot use a string pattern on a bytes-like object
]]]

I guess that's one of the issues fixed by the swig-py3 branch.

- Then, on the swig-py3 branch, with py 3.7.4 I ran into this issue:
https://bugs.python.org/issue37549 (os.dup() fails for standard
streams on Windows 7)
This fails for any *.py test, because of line 836 in build/run_tests.py:
        old_stdout = os.dup(sys.stdout.fileno())
It errors out with:
       OSError: [WinError 87] The parameter is incorrect

- Upgraded to py 3.7.5, in which the above issue seems to be fixed.
Now, *.py tests still don't work. I get no output at all:
[[[
C:\research\svn\dev\swig-py3>python win-tests.py --release -t tree_conflict .
'ruby' is not recognized as an internal or external command,
operable program or batch file.
Testing Release configuration on local repository.
[1/1] tree_conflict_tests.py
C:\research\svn\dev\swig-py3>
]]]

tests.log only contains one line:
    START: tree_conflict_tests.py

However, if I run it with --log-to-stdout, the tests do work (with a
lot of output on stdout). I.e. I get some fails without your
fix_tree_conflict_tests_patch.txt, and all tests successful if I apply
the patch.


Conclusion: I can confirm your patch works on Windows, for both Pyton
2.7.16 and 3.7.5 on the swig-py3 branch. As for the stdout
redirection, I guess there might still be a problem ... perhaps the
fix for https://bugs.python.org/issue37549 is not sufficient for
Windows 7 ... dunno. Maybe someone can try this on Windows 10 and see
if it makes a difference.

-- 
Johan

Mime
View raw message