RE: Re: question clarification needed (tagging with embedded externals)

was hoping to avoid cross posting but - here is what the subversion team had to say about this issue, im sorry but its a bit lengthly. has anyone else here had to deal with this issue (taggng & externals)?

--- On Mon, 7/20/09, Ryan Schmidt
Date: Monday, July 20, 2009, 2:26 PM

On Jul 20, 2009, at 15:25, steve wrote:

> Ok I came across a situation using tortoise and SVN
>
> (http://tortoisesvn.tigris.org/ds/viewMessage.do?dsForumId=4061&dsMessageId=2371888)
>
> Basically:
> Ok so i was told to:
>
>> Google for svncopy.pl. It is a Perl script that works with Subversion
>> and is intended to do what I think you are trying to accomplish.
>
> And from my understanding this script would allow me to "pin down" a external for a 'tag' operation is that right?

That's the idea. Any external in a thing you copy (to a new tag or to anywhere else) will get the latest revision burned into the external definition, so that if you later check out that copy (e.g. that tag) you will get the revision of the external that was in effect at the time the copy (tag) was made, as opposed to the revision of the external at the time you check out, which is what you would get otherwise.

The other option is to always put a revision specifier in every external, even on trunk. This is in fact a good idea, IMHO. Because you don't just care about keeping the external in sync when checking out a tag. You also care when checking out an older version of trunk. In one of my projects, I have Program A which has an external pointing to a specific revision (let's say r123) of Library B. This means I can commit any changes I want to Library B, including ones that change its API, without fear of breaking Program A. Later, when I am in the mood to work on Program A, I can update the Library B external to a later revision (let's say r146) and make the necessary changes to Program A so that it can use the new API of Library B.

If you want to have a look, "Program A" is canviz.js, "Library B" is path.js, and the repository is available by following links from http://canviz.org/

> um - does anyone know if there is a way to run this as a hook through the tortoise windows gui?

Subversion has server-side hooks; it does not have client-side hooks, and probably won't for a long time or ever for a variety of reasons which have been explored in many previous mailing list posts.

The folks on the TortoiseSVN mailing list are better-equipped to answer TortoiseSVN-specific questions. I did a quick Google search and found that TortoiseSVN does in fact have client-side hooks. They are documented here:

http://tortoisesvn.net/docs/release/TortoiseSVN_en/tsvn-dug-settings.html#tsvn-dug-settings-hooks

I don't know if they can be used to implement what you want. I do not use TortoiseSVN.

The developers of TortoiseSVN could perhaps take the ideas of svncopy.pl and integrate them into TortoiseSVN and give you an option to use that, but since you were referred to svncopy.pl by the TortoiseSVN mailing list, I guess this has not thus far occurred.

> Is there a way to replace the whole existing 'svn copy' command on the server with this new version?

You would use svncopy.pl instead of using "svn copy" or Tortoise SVN's copy command.

> Which leds me to wonder after a bit more looking it seems like this same issue might have been addressed as part of "Revision 38450" (?)

The log message of revision 38450 in the Subversion repository reads as follows; I'm not sure if or how this revision relates to this issue.

$ svn log -r 38450 http://svn.collab.net/repos/svn
------------------------------------------------------------------------
r38450 | hwright | 2009-07-20 13:03:36 -0500 (Mon, 20 Jul 2009) | 34 lines

Rev another libsvn_wc private-public API.

(As an aside, having these APIs feels a bit like cheating. I mean, if we use
them, and they are useful, shouldn't we expose them to the rest of the world?
Not only do we let others use these APIs, but we also get rid of this
"svn_wc__blah() is not really internal, so we need an *additional* wrapper
around it to pull out the wc_db param" dance. Oh well, maybe in some future
life...)

* subversion/include/private/svn_wc_private.h
  (svn_wc__versioned_file_modcheck): Rev, adding a wc_ctx, and updating the
    docstring.

* subversion/libsvn_wc/wc.h
  (svn_wc__internal_versioned_file_modcheck): New.

* subversion/libsvn_wc/update_editor.c
  (merge_file): Update caller.

* subversion/libsvn_wc/questions.c
  (compare_and_verify): Get rid of the access baton parameter, and use a
    wc_db instead. Accept only absolute paths.
  (svn_wc__internal_versioned_file_modcheck): New.
  (svn_wc__versioned_file_modcheck): Update args, call the wrapper.
  (svn_wc__text_modified_internal_p): Get a db and absolute paths to call
    compare_and_verify().

* subversion/libsvn_wc/log.c
  (log_do_committed): Update caller.

* subversion/libsvn_client/merge.c
  (files_same_p): Remove the adm_access param, and require absolute paths.
  (merge_file_deleted): Get some absolute paths and update a caller.

------------------------------------------------------------------------

> and guess i was really looking for more info - wondering if there has indeed been a patch or something of this sort to address this issue?

As far as I know, Subversion behaves as intended, so there is nothing to patch, no issue to address, except perhaps in terms of making the documentation clearer about the behavior that Subversion exhibits.

------------------------------------------------------
http://tortoisesvn.tigris.org/ds/viewMessage.do?dsForumId=4061&dsMessageId=2374465

To unsubscribe from this discussion, e-mail: [users-unsubscribe_at_tortoisesvn.tigris.org].
Received on 2009-07-22 17:04:56 CEST