subversion-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Greg Stein <gst...@gmail.com>
Subject Re: Coniguring 301/302 redirects to track an fspath rename
Date Tue, 05 Feb 2013 03:57:15 GMT
Better handling of redirects was released in 1.7.0.

See: http://subversion.tigris.org/issues/show_bug.cgi?id=2779 and the
CHANGES file.

The working copy is not relocated, however. Just the error report. The
user needs to follow up with some action. Other clients (eg
TortoiseSVN) could theoretically perform an auto relocation, but the
cmdline client does not.

Cheers,
-g


On Mon, Feb 4, 2013 at 10:41 PM, Joe Schaefer <joe_schaefer@yahoo.com> wrote:
> Has Daniel been pulling my leg this whole time
> about this automatic relocation feature?  I don't
> see any mention of it on the website for the 1.7
> docs and have tested through 1.7.3 without seeing
> it work.
>
> If there's work here that needs doing on
> the subversion side I apologize for all the noise
> I've generated, I thought this was a feature you
> were already supporting on some level and was anxious
> to take advantage of it.
>
>
> ________________________________
> From: Joe Schaefer <joe_schaefer@yahoo.com>
> To: Greg Stein <gstein@gmail.com>
> Cc: "dev@subversion.apache.org" <dev@subversion.apache.org>
> Sent: Monday, February 4, 2013 8:15 PM
>
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> Ok folks, some butt ugly code I wrote is fishing
> the source url out of the REPORT request body on
> update requests.  If I checkout revision 1400000
> of
>
> http://svn.apache.org/repos/asf/incubator/openmeetings/trunk/singlewebapp
>
> that works fine, but then when I try to run
>
> $ svn up
>
> on that working copy, I get the following error:
>
>   svn: E175011: Repository moved permanently to
>   'https://svn.apache.org/repos/asf/openmeetings/trunk/singlewebapp';
>   please relocate
>
> Honestly that's an improvement, but not what I'm looking
> for here: I need a way for the client to autoswitch to the
> new location, not merely prompt the user to do it manually.
> This is with svn 1.7.0, maybe things have changed since then?
>
>
> ________________________________
> From: Joe Schaefer <joe_schaefer@yahoo.com>
> To: Greg Stein <gstein@gmail.com>
> Cc: "dev@subversion.apache.org" <dev@subversion.apache.org>
> Sent: Monday, February 4, 2013 4:48 PM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> Damn, processing updates requires fishing little
> details out of a REPORT request body.  This is getting
> complicated...
>
>
> ________________________________
> From: Joe Schaefer <joe_schaefer@yahoo.com>
> To: Greg Stein <gstein@gmail.com>
> Cc: "dev@subversion.apache.org" <dev@subversion.apache.org>
> Sent: Monday, February 4, 2013 3:57 PM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> Here's the new section for the additional PROPFIND
> rewrite block:
>
> /* RewriteCond %{SUBREQ} =false
> */
> /* RewriteCond %{REQUEST_METHOD} =PROPFIND
> */
> /* RewriteCond %{REQUEST_URI}
> /repos/asf/!svn/(?:bc|rvr)/(\d+)/incubator/(.+)*/
> /* RewriteCond /repos/asf/%2?ckpath  -U
> */
> /* RewriteRule .* /repos/asf/%2 [L,R=301]
> */
>
>
> As I said it works for me with openmeetings checkouts
> out of incubator (1.6 and 1.7 clients point me at the
> new location).  I am leery tho about dorking with !svn
> based urls but these !svn/bc/$revision and !svn/rvr/$revision
> ones are nice candidates because of the embedded revision
> and path information that leads me to think I'm not too
> far off in guessing that without interception by my module
> we'd just be serving up 404's for them when they are looking
> in the incubator tree for graduated podling resources.
>
>
> ________________________________
> From: Joe Schaefer <joe_schaefer@yahoo.com>
> To: Greg Stein <gstein@gmail.com>
> Cc: "dev@subversion.apache.org" <dev@subversion.apache.org>
> Sent: Monday, February 4, 2013 2:44 PM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> Mixed news- I hacked together PROPFIND support
> for my module and it's doing the "right thing"
> on checkouts with 1.6 and 1.7. NFI what else this
> breaks, but if folks could beat on incubating repos
> for a bit to see I'd appreciate it.
>
> Thanks!
>
>
> ________________________________
> From: Joe Schaefer <joe_schaefer@yahoo.com>
> To: Greg Stein <gstein@gmail.com>
> Cc: "dev@subversion.apache.org" <dev@subversion.apache.org>
> Sent: Monday, February 4, 2013 1:46 PM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> Oh damn, sigh.  I'm completely wrong- subversion doesn't
> serve a 404 on the OPTIONS request, even for a resource
> that's not present in HEAD.  It serves a 200 response,
> and only several PROPFIND's later do we actually serve
> a showstopping 404 on a checkout for a non-existent resource.
>
> That complicates this situation significantly, sorry.
> I don't see an immediate way to satisfy the ASF's use
> case given the current svn support for redirects revolves
> around the initial OPTIONS response not PROPFINDS.
>
>
> ________________________________
> From: Joe Schaefer <joe_schaefer@yahoo.com>
> To: Greg Stein <gstein@gmail.com>
> Cc: "dev@subversion.apache.org" <dev@subversion.apache.org>
> Sent: Monday, February 4, 2013 1:30 PM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> Please correct me if I'm mistaken, but my impression
> from looking at server logs is that there are certain
> cases where a 404 response to an initial OPTIONS request
> is a showstopper (like a checkout or update operation),
> but other times the 404 response is not a problem (like
> for operations involving pegrevs).  If we can identify
> this distinction and flag it somehow in the initial request
> so that in the first case we ask for a non-showstopper
> response from the server if available, that's all I need
> to support the ASF's use case in my module.  Providing
> additional version info in the query string can refine
> the server response my httpd module would provide, but
> that aspect is not essential to service the ASF's needs.
>
>
> ________________________________
> From: Joe Schaefer <joe_schaefer@yahoo.com>
> To: Greg Stein <gstein@gmail.com>
> Cc: "dev@subversion.apache.org" <dev@subversion.apache.org>
> Sent: Monday, February 4, 2013 12:26 PM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> How about a "prefer redirect" flag somewhere in
> the OPTIONS request?  Anything to signal to the
> server that the request wants something better
> than a 404 response it possible.
>
>
> ________________________________
> From: Greg Stein <gstein@gmail.com>
> To: Joe Schaefer <joe_schaefer@yahoo.com>
> Cc: dev@subversion.apache.org
> Sent: Monday, February 4, 2013 7:44 AM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> Yeah... I think the best we can do in the OPTIONS is to hint. It can't
> really be more than a hint, since a connection can be used for N various
> operations.
> What might be more ideal is to respect redirects after that initial OPTIONS.
> Then your module could just wait until it knows the right data.
> It's not an easy problem. If we get a 301, then we should definitely rewrite
> the working copy to use the new URL. But... is a 301 correct, if "old"
> revisions still work? ....
> Cheers,
> -g
> On Feb 3, 2013 2:33 PM, "Joe Schaefer" <joe_schaefer@yahoo.com> wrote:
>
> We don't actually need the subversion server
> to do anything useful with the query string
> on an OPTIONS request.  We just need it as
> a hint for http server admins, combined with
> a (global) configuration option to tell our
> working copies to autoswitch when they see a
> 301 on the initial OPTIONS response.
>
> Sorry if this is getting boring.  In no way is
> it essential, but we have significant working copy
> investments in incubating projects that are a pain
> to manage administratively right now whenever a project
> graduates.  Anything along these lines to cut down
> on the human labor aspects would be most welcome.
>
>
> ________________________________
> From: Joe Schaefer <joe_schaefer@yahoo.com>
> To: Greg Stein <gstein@gmail.com>
> Cc: "dev@subversion.apache.org" <dev@subversion.apache.org>
> Sent: Sunday, February 3, 2013 1:45 PM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> IOW the easiest way for us to support OPTIONS
> requests with this module, which would require
> no code changes to the module I wrote, is for the
> initial OPTIONS request to hang a ?p=whatever
> query string onto the request whenever it's dealing
> with pegrevs.
>
>
>
> ________________________________
> From: Joe Schaefer <joe_schaefer@yahoo.com>
> To: Greg Stein <gstein@gmail.com>
> Cc: "dev@subversion.apache.org" <dev@subversion.apache.org>
> Sent: Sunday, February 3, 2013 1:04 PM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> A few urls to fetch to get the basic idea:
>
> http://svn.apache.org/repos/asf/incubator/openmeetings/trunk
> (will redirect)
>
> http://svn.apache.org/repos/asf/incubator/openmeetings/trunk?p=1400000
> (won't redirect)
>
>
>
> ________________________________
> From: Joe Schaefer <joe_schaefer@yahoo.com>
> To: Greg Stein <gstein@gmail.com>
> Cc: "dev@subversion.apache.org" <dev@subversion.apache.org>
> Sent: Sunday, February 3, 2013 12:15 PM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> The code is here if you are interested (dunno
> if it's publicly accessible or not but I believe
> so).
>
> https://svn.apache.org/repos/infra/infrastructure/trunk/projects/svn_check_path/
>
> ________________________________
> From: Joe Schaefer <joe_schaefer@yahoo.com>
> To: Greg Stein <gstein@gmail.com>
> Cc: "dev@subversion.apache.org" <dev@subversion.apache.org>
> Sent: Sunday, February 3, 2013 12:10 PM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> I backed off with the idea of trying to coax
> new behavior out of svn clients by munging OPTIONS
> responses, but GET and HEAD are now doing
> interesting things even with ?p= set.  For our
> purposes someday it would be nice to just have
> a graduating podling's checkouts all auto-switch
> to the moved location.  Whether that's
> accomplished by augmenting an OPTIONS request
> with revision details or just using the existing
> GET/HEAD support or some other method would be
> just fine with me.
>
>
> ________________________________
> From: Joe Schaefer <joe_schaefer@yahoo.com>
> To: Greg Stein <gstein@gmail.com>
> Cc: "dev@subversion.apache.org" <dev@subversion.apache.org>
> Sent: Saturday, February 2, 2013 2:07 AM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> The (pegrev) revision is typically available on
> the command-line: all I want is for svn to distinguish
> between
>
>     svn ls foo@1000000
>
> and
>
>     svn ls foo
>
> as far as making redirects pegrev-aware.
>
>
> ________________________________
> From: Greg Stein <gstein@gmail.com>
> To: Joe Schaefer <joe_schaefer@yahoo.com>
> Cc: dev@subversion.apache.org
> Sent: Saturday, February 2, 2013 1:53 AM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> Oh, and to the second part: the code which sends OPTIONS has no knowledge
> about the future operations. There is no way to send <revision/>, or
> similar. *Very* disconnected, as in: not even cheesy-hackable.
> Cheers,
> -g
> On Feb 2, 2013 1:49 AM, "Greg Stein" <gstein@gmail.com> wrote:
>
> That OPTIONS request is generic, and contains no information about the
> future operation(s) that will be performed on the connection. It is used for
> the client to validate and gather information about the server.
> Cheers,
> -g
> On Feb 2, 2013 1:23 AM, "Joe Schaefer" <joe_schaefer@yahoo.com> wrote:
>
> So I now see the request body (xml payload)
> in the OPTIONS request, but nothing there nor
> in the headers tells me about revision numbers.
> If I can convince you to add a <revision/> block
> to the OPTIONS request body I can handle the rest
> from the httpd side.  Obviously this won't help
> existing clients, but hey it's a gee-whiz feature
> anyhow.
>
>
>
> ________________________________
> From: Joe Schaefer <joe_schaefer@yahoo.com>
> To: 'Daniel Shahaf' <d.s@daniel.shahaf.name>; Bert Huijben <bert@qqmail.nl>
> Cc: "dev@subversion.apache.org" <dev@subversion.apache.org>
> Sent: Friday, February 1, 2013 9:26 PM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> So I have this implemented about as well
> as I can with what I know about OPTIONS
> requests that svn generates.  It would
> help if I knew how svn supplies revision
> information in the OPTIONS request headers
> so I can pass that along to the codebase
> instead of always using the youngest rev.
>
>
> ________________________________
> From: 'Daniel Shahaf' <d.s@daniel.shahaf.name>
> To: Bert Huijben <bert@qqmail.nl>
> Cc: dev@subversion.apache.org
> Sent: Friday, February 1, 2013 3:33 PM
> Subject: Re: Coniguring 301/302 redirects to track an fspath rename
>
> Bert Huijben wrote on Fri, Feb 01, 2013 at 21:28:10 +0100:
>>
>>
>> > -----Original Message-----
>> > From: Daniel Shahaf [mailto:d.s@daniel.shahaf.name]
>> > Sent: vrijdag 1 februari 2013 19:11
>> > To: dev@subversion.apache.org
>> > Subject: Coniguring 301/302 redirects to track an fspath rename
>> >
>> > Does anyone have an example of how to configure a server to issue
>> > 301/302 redirects for an fspath that had been renamed?
>> >
>> > For example we have
>> >
>> >    <Location /repos/asf>
>> >    SVNPath ...
>> >    </Location>
>> >
>> > And we'd like to do:
>> >
>> >    # The project was renamed
>> >    Redirect /repos/asf/openejb https://svn.apache.org/repos/asf/tomee
>> >
>> > but we're hitting various problems:
>> >
>> > - The redirect kicks in for historical revisions (prior to the 'svn mv
>> >  ^/openejb ^/tomee' in r1432805) too, such as:
>> >  https://svn.apache.org/repos/asf/openejb?p=1400000
>> >
>> > - A similar configuration failed to kick in during update/checkout of
>> >  working copy checked out from (a pre-rename revision of) ^/openejb:
>> >  the initial request got matched and redirected, but a subsequent
>> >  request to /repos/asf/!svn/.../openejb failed to match.
>> >
>> > Ideally we'd like to issue a 301 redirect for requests to /openejb that
>> > concern r1432805 or later, but leave requests concerning r1432804 or
>> > earlier untouched.
>> >
>> > (or maybe what we *really* want is a repos-side symlink... but we're
>> > running 1.7, not 1.9, and we'll appreciate solutions that work within
>> > that limitation :))
>>
>> We currently only support redirects above the repository level.
>>
>> Redirections inside would be a completely different feature.
>>
>
> OK... :(
>
>> Why not just leave a top level folder with some readme?
>>
>
> Every time a podling graduates from the incubator, we do a rename:
>   svn mv ^/incubator/flex ^/flex
>
> If we can return 301 whenever somebody does 'svn up' in a wc of
> ^/incubator/flex, we'll save many users (2-4 projects every month)
> having to learn about 'svn relocate'.
>
>> I think you should be able to redirect the normal webbrowser GETs though,
>> as
>> I don't think we use those urls from our ra layers. (Or did we start using
>> them for HEAD requests in HTTPv2?)
>>
>
> 'svn ls $URL@peg' was affected by the redirects I had.
>
>>     Bert
>>
>
> Thanks,
>
> Daniel
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>

Mime
View raw message