Re: [PATCH] allow unversioned parent for copy src again

To the extent that I'm competent to comment (not much!) that looks like
what I was considering trying on my own.

Thank you for the fix.

Karl Fogel wrote:
> Chris Rose <chris.rose_at_messagingdirect.com> writes:
>> I'm sorry to prod on this, but I'm quite concerned about the change in
>> functionality, and would like to find out if this will remain in place
>> for 1.5, or if I should register it as a bug with the client?
>>
>> Can somebody weigh in either way, please?
>
> It sounds like a bug to me. I've filed an issue for it, and written a
> tentative patch. Here's the issue:
>
> http://subversion.tigris.org/issues/show_bug.cgi?id=3068
>
> The patch is below. I'd love some other eyes on it before I commit,
> though, from people more familiar with wc mergeinfo propagation.
>
> Thanks for the reproduction script, btw, that helped a lot.
>
> Here's the patch:
>
> [[[
> Fix issue #3068: wc->wc copy should work when src parent is unversioned.
>
> * subversion/libsvn_client/copy.c
> (do_wc_to_wc_copies): Propagate mergeinfo from parent only if possible.
>
> * subversion/libsvn_wc/lock.c
> (do_open): Document an error return that callers depend on.
> ]]]
>
> Index: subversion/libsvn_client/copy.c
> ===================================================================
> --- subversion/libsvn_client/copy.c (revision 28822)
> +++ subversion/libsvn_client/copy.c (working copy)
> @@ -347,10 +347,18 @@
> }
> else
> {
> - SVN_ERR(svn_wc_adm_open3(&src_access, NULL, src_parent, FALSE,
> - pair->src_kind == svn_node_dir ? -1 : 0,
> - ctx->cancel_func, ctx->cancel_baton,
> - iterpool));
> + err = svn_wc_adm_open3(&src_access, NULL, src_parent, FALSE,
> + pair->src_kind == svn_node_dir ? -1 : 0,
> + ctx->cancel_func, ctx->cancel_baton,
> + iterpool);
> + /* The parent of a copy src might not be versioned at all. */
> + if (err && err->apr_err == SVN_ERR_WC_NOT_DIRECTORY)
> + {
> + src_access = NULL;
> + svn_error_clear(err);
> + err = NULL;
> + }
> + SVN_ERR(err);
> }
>
> /* Perform the copy */
> @@ -365,12 +373,13 @@
> if (err)
> break;
>
> - err = propagate_mergeinfo_within_wc(pair, src_access, dst_access,
> - ctx, pool);
> + if (src_access)
> + err = propagate_mergeinfo_within_wc(pair, src_access, dst_access,
> + ctx, pool);
> if (err)
> break;
>
> - if (src_access != dst_access)
> + if (src_access && src_access != dst_access)
> SVN_ERR(svn_wc_adm_close(src_access));
> }
>
> Index: subversion/libsvn_wc/lock.c
> ===================================================================
> --- subversion/libsvn_wc/lock.c (revision 28822)
> +++ subversion/libsvn_wc/lock.c (working copy)
> @@ -542,6 +542,9 @@
> /* This is essentially the guts of svn_wc_adm_open3, with the additional
> * parameter UNDER_CONSTRUCTION that gets set TRUE only when locking the
> * admin directory during initial creation.
> + *
> + * If the working copy is already locked, return SVN_ERR_WC_LOCKED; if
> + * it is not a versioned directory, return SVN_ERR_WC_NOT_DIRECTORY.
> */
> static svn_error_t *
> do_open(svn_wc_adm_access_t **adm_access,

-- 
Chris Rose
Developer    Planet Consulting Group
(780) 577-8433
crose_at_planetci.com