> -----Original Message-----
> From: Ivan Zhakov [mailto:ivan_at_visualsvn.com]
> Sent: maandag 9 februari 2015 14:00
> To: Bert Huijben
> Cc: Subversion Development
> Subject: Re: svn commit: r1654810 -
> /subversion/trunk/subversion/tests/cmdline/prop_tests.py
>
> On 9 February 2015 at 15:54, Bert Huijben <bert_at_qqmail.nl> wrote:
> >
> >
> >> -----Original Message-----
> >> From: Ivan Zhakov [mailto:ivan_at_visualsvn.com]
> >> Sent: maandag 9 februari 2015 13:16
> >> To: Subversion Development
> >> Cc: Bert Huijben
> >> Subject: Re: svn commit: r1654810 -
> >> /subversion/trunk/subversion/tests/cmdline/prop_tests.py
> >>
> >> On 26 January 2015 at 17:17, <rhuijben_at_apache.org> wrote:
> >> > Author: rhuijben
> >> > Date: Mon Jan 26 14:17:25 2015
> >> > New Revision: 1654810
> >> >
> >> > URL: http://svn.apache.org/r1654810
> >> > Log:
> >> > Following up on r1654794, add a related test that verifies the author value
> >> > retrieved with the stat ra functions.
> >> >
> >> > * subversion/tests/cmdline/prop_tests.py
> >> > (xml_unsafe_author2): New test.
> >> > (test_list): Add xml_unsafe_author2.
> >> >
> >> > Modified:
> >> > subversion/trunk/subversion/tests/cmdline/prop_tests.py
> >> >
> >> > Modified: subversion/trunk/subversion/tests/cmdline/prop_tests.py
> >> > URL:
> >>
> http://svn.apache.org/viewvc/subversion/trunk/subversion/tests/cmdline/prop_
> >> tests.py?rev=1654810&r1=1654809&r2=1654810&view=diff
> >> >
> >>
> ================================================================
> >> ==============
> >> > --- subversion/trunk/subversion/tests/cmdline/prop_tests.py (original)
> >> > +++ subversion/trunk/subversion/tests/cmdline/prop_tests.py Mon Jan 26
> >> 14:17:25 2015
> >> > @@ -2669,6 +2669,46 @@ def xml_unsafe_author(sbox):
> >> > 'proplist', '--revprop', '-r', '1', '-v',
> >> > wc_dir)
> >> >
> >> > +@Issue(4415)
> >> > +def xml_unsafe_author2(sbox):
> >> > + "svn:author with XML unsafe chars 2"
> >> > +
> >> > + sbox.build(create_wc = False)
> >> > + repo_url = sbox.repo_url
> >> > +
> >> > + svntest.actions.enable_revprop_changes(sbox.repo_dir)
> >> > +
> >> > + # client sends svn:author (via PROPPATCH for DAV)
> >> > + svntest.actions.run_and_verify_svn(None, None, [],
> >> > + 'propset', '--revprop', '-r', '1',
> >> > + 'svn:author', 'foo\bbar', repo_url)
> >> > +
> >> > + # Ensure a stable date
> >> > + svntest.actions.run_and_verify_svn(None, None, [],
> >> > + 'propset', '--revprop', '-r', '1',
> >> > + 'svn:date', '2000-01-01T12:00:00.0Z',
> >> > + repo_url)
> >> > +
> >> > + if svntest.main.is_ra_type_dav():
> >> > + # This receives the filtered author (but that is better than an Xml fail)
> >> > + expected_author = 'foobar'
> >> > + else:
> >> > + expected_author = 'foo\bbar'
> >> > +
> >> > + expected_output = svntest.verify.UnorderedOutput([
> >> > + ' 1 %-8s Jan 01 2000 ./\n' % expected_author,
> >> > + ' 1 %-8s Jan 01 2000 A/\n' % expected_author,
> >> > + ' 1 %-8s 25 Jan 01 2000 iota\n' % expected_author
> >> > + ])
> >> Hi Bert,
> >>
> >> This test is failing on system with non-US default locale, because
> >> 'svn ls -v' prints date in locale depended format.
> >
> > I run my system with Dutch date formats but still see the US format when
> running the tests (because we force a locale somewhere), while a normal
> invocation of 'svn -v' does show it in a different format.
> > (In my case 'Jan' would just be 'jan' (different capitalization))
> >
> > Are you sure that there isn't a different problem here?
> >
> > The test can probably be tweaked to use some kind of regex matching, but I
> doubt whether that is necessary.
> >
> There are several settings in Windows that affect locale used for
> applications. Do you have configured "Language for non-Unicode
> programs" to Dutch on your system?
We use apr_strftime() to format the sting and it behaves different when I run it in the test suite and normally.
I would be very surprised if this function would use *multiple different* locale settings at the same time.
If that is the case we might have completely different problems that we should look at. It appears we also assume that this function always delivers proper utf-8 strings, which after looking through the source of both apr and the Visual C++ CRT sourcecode that implements strftime think is a bad assumption.
A few years ago we added code to the test suite that should force the locale to the expected locale and that works for me...
Before further ignoring the problem by just making the test pass I think we should look at that.
Bert
Received on 2015-02-09 14:14:57 CET