Daniel Shahaf <danielsh_at_apache.org> writes:
> On Wed, Sep 04, 2013 at 11:50:49PM -0400, James McCoy wrote:
>> This was with svn 1.7.9. I haven't had a chance to test newer
>> versions, but looking through the code in trunk, it looks like it
>> would still be a problem.
> Confirmed, it reproduces in trunk:
> Index: subversion/tests/cmdline/upgrade_tests.py
> --- subversion/tests/cmdline/upgrade_tests.py (revision 1520363)
> +++ subversion/tests/cmdline/upgrade_tests.py (working copy)
> @@ -425,7 +425,7 @@ def simple_entries_replace(path, from_url, to_url)
> def basic_upgrade_1_0(sbox):
> "test upgrading a working copy created with 1.0.0"
> - sbox.build(create_wc = False)
> + sbox.build(name='foo+', create_wc = False)
> replace_sbox_with_tarfile(sbox, 'upgrade_1_0.tar.bz2')
> url = sbox.repo_url
> % ../runpytest upgrade basic_upgrade_1_0 2>&1 | head
> W: lt-svn: subversion/libsvn_subr/path.c:119: svn_path_join_internal: Assertion `svn_path_is_canonical_internal(base, pool)' failed.
> W: CMD: /home/danielsh/src/svn/t1/subversion/svn/svn upgrade 'svn-test-work/working_copies/foo+' --config-dir /home/danielsh/src/svn/t1/subversion/tests/cmdline/svn-test-work/local_tmp/config --password rayjandom --no-auth-cache --username jrandom terminated by signal 6
#0 0x00007ffff72cc475 in *__GI_raise (sig=<optimized out>)
#1 0x00007ffff72cf6f0 in *__GI_abort () at abort.c:92
#2 0x00007ffff72c5621 in *__GI___assert_fail (
assertion=0x7ffff7d49210 "svn_path_is_canonical_internal(base, pool)",
file=<optimized out>, line=119,
function=0x7ffff7d498f0 "svn_path_join_internal") at assert.c:81
#3 0x00007ffff7d14a2e in svn_path_join_internal (
base=0x7ffff3bc4bc0 "file:///home/pm/sw/subversion/obj/subversion/tests/cmdline/svn-test-work/repositories/foo%2B/trunk", component=0x7ffff3bc56d0 "iota",
pool=0x7ffff7e3e028) at ../src/subversion/libsvn_subr/path.c:119
#4 0x00007ffff7d16770 in svn_path_url_add_component2 (
url=0x7ffff3bc4bc0 "file:///home/pm/sw/subversion/obj/subversion/tests/cmdline/svn-test-work/repositories/foo%2B/trunk", component=0x7ffff3bc56d0 "iota",
pool=0x7ffff7e3e028) at ../src/subversion/libsvn_subr/path.c:1085
It looks like the %2B is failing svn_uri_is_canonical. Looking at
svn_uri__char_validity in libsvn_subr/path.c the table says that '+'
should not be % encoded. The problem is that the Python testsuite uses
urllib.pathname2url(self.repo_dir) and that has different ideas, it
does % encode '+'.
Philip Martin | Subversion Committer
WANdisco // *Non-Stop Data*
Received on 2013-09-05 20:52:56 CEST