[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: [PATCH] Canonicalize dirent/URL before passing to client API

From: Julian Foad <julian.foad_at_wandisco.com>
Date: Thu, 11 Nov 2010 16:43:44 +0000

On Wed, 2010-11-10, Noorul Islam K M wrote:
> [[[
>
> Canonicalize dirent/URL before passing to client API
>
> * subversion/svn/merge-cmd.c (svn_cl__merge),
> subversion/svn/propget-cmd.c (svn_cl__propget),
> subversion/svn/proplist-cmd.c (svn_cl__proplist),
> subversion/svn/copy-cmd.c (svn_cl__copy),
> subversion/svn/mergeinfo-cmd.c (svn_cl__mergeinfo),
> subversion/svn/blame-cmd.c (svn_cl__blame),
> subversion/svn/log-cmd.c (svn_cl__log),
> subversion/svn/export-cmd.c (svn_cl__export),
> subversion/svn/info-cmd.c (svn_cl__info):
> Use svn_cl__opt_parse_path() instead of svn_opt_parse_path()
> followed by canonicalizing.

I thought that part looked good... but then discovered something. See
below.

> * subversion/svn/propdel-cmd.c (svn_cl__propdel),
> subversion/svn/propget-cmd.c (svn_cl__propget),
> subversion/svn/propset-cmd.c (svn_cl__propset),
> subversion/svn/proplist-cmd.c (svn_cl__proplist):
> Canonicalize URL

Let's do that part inside the function svn_cl__revprop_prepare()
instead, like I mentioned in another email, and not include it in this
patch. (You missed the call in propedit-cmd, and it won't be possible
to miss any calls if we do it that way.)

> * subversion/tests/cmdline/mergeinfo_tests.py
> (mergeinfo_url_special_characters),
> subversion/tests/cmdline/prop_tests.py
> (props_url_special_characters),
> subversion/tests/cmdline/merge_tests.py
> (merge_url_special_characters),
> subversion/tests/cmdline/log_tests.py
> (log_url_special_characters),
> subversion/tests/cmdline/copy_tests.py
> (copy_url_special_characters),
> subversion/tests/cmdline/blame_tests.py
> (blame_url_special_characters):
> New tests

The purpose of the tests is to check that the commands properly accept
and canonicalize non-canonical target paths. That's not clear from just
reading them, so let's add a comment in each one just to say that, so
that future maintainers will know. It might be a good idea to rename
the test functions and description away from "special characters" to
something like "non-canonical targets", since a path or URL can be
non-canonical even if it doesn't contain "special" characters.

Some of these subcommands accept either a local path or URL, but the
tests only check URLs - was that intentional?

> +def mergeinfo_url_special_characters(sbox):
> + """special characters in svn mergeinfo URL"""
> +
> + sbox.build()
> + wc_dir = sbox.wc_dir
> + special_url = sbox.repo_url + '/%2E'

Why construct a URL that is non-canonical in two ways at the same time?
The fact that a character that doesn't need to be percent-encoded ('.'
in this case) is percent-encoded makes it non-canonical, but ending a
URL with '/.' also makes it non-canonical. You should be able to test
for each of those separately.

Aha. I think I have discovered the problem. The arguments have already
been through svn_uri_canonicalize() once, inside
svn_cl__args_to_target_array_print_reserved(). The trouble is that it
didn't work properly it converted

  ".../%2E" to ".../."

but it should have converted

  ".../%2E" to "..."

I think. Can a URI expert confirm this?

So I think running the output through svn_uri_canonicalize again is not
the right solution, and we need to fix svn_uri_canonicalize instead.

If that analysis is right, then we will indeed need to test a URL like
repo_url + '/%2e', but only a single test.

- Julian
Received on 2010-11-11 17:44:25 CET

This is an archived mail posted to the Subversion Dev mailing list.