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

Re: Cleaning up XFails: locks-test.exe 9 and 10

From: C. Michael Pilato <cmpilato_at_collab.net>
Date: Mon, 07 Feb 2011 16:14:16 -0500

On 02/07/2011 03:44 PM, Philip Martin wrote:
> Paul Burba <ptburba_at_gmail.com> writes:
>
>> Do "DAV clients sometimes LOCK non-existent paths, as a way of
>> reserving names"? I'm not sure exactly what that means, does anyone
>> have an inkling?
>
> In subversion/mod_dav_svn/lock.c:append_locks the code does a PUT of a
> 0-byte file when an attempt is made to lock a non-existant file. You
> need SVNAutoversioning enabled and you need to be using a non-Subversion
> client. It is possible to trigger this using cadaver:
>
> $ svnadmin create repo
> $ ln -sf /bin/true repo/hooks/pre-lock
> $ cadaver http://localhost:8888/obj/repo
> dav:/obj/repo/> lock f
> Locking `f': succeeded.
> dav:/obj/repo/> quit
> $ svnlook tree repo
> /
> f

While generic DAV clients can "LOCK non-existent paths", Subversion's
behavior is to immediate turn those into non-non-existing paths. :-) We
decided long ago that we didn't want to support the so-called "LOCK-NULL
resources" (locks on non-existent paths), so the code cheats to avoid this.
 As such, the tests which attempt to validate that we do support these can
be tossed, in my opinion.

And I'd say you could toss the directory locking test, too. The problem
space was large enough that we bailed on supporting it in the first release
of the locking support. So if we ever do decide to support this, we'll have
to design a full suite of tests with coverage that's advanced enough to
match the complexity of the problem space.

-- 
C. Michael Pilato <cmpilato_at_collab.net>
CollabNet   <>   www.collab.net   <>   Distributed Development On Demand

Received on 2011-02-07 22:15:00 CET

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