On Jun 29, 2012 6:05 PM, <rhuij...@apache.org> wrote:
>...
> +/* Deserialize a svn_wc_conflict_version_t from the skel.
> + Set *LOCATION to NULL when the data is not a svn_wc_conflict_version_t. */
> +static svn_error_t *
> +conflict__read_location(const svn_wc_conflict_version_t **location,
> + const svn_skel_t *skel,
> + apr_pool_t *result_pool,
> + apr_pool_t *scratch_pool)
> +{
> + const char *repos_root_url;
> + const char *repos_uuid;
> + const char *repos_relpath;
> + svn_revnum_t revision;
> + apr_int64_t v;
> + svn_node_kind_t node_kind; /* note that 'none' is a legitimate value */
Please (de)serialize using svn_kind_t. I recognize that
conflict_version_t uses the old kind; just convert. But it would be
great to serialize with the new type.
> + const char *kind_str;
> +
> + svn_skel_t *c = skel->children;
const
> +
> + if (!svn_skel__matches_atom(c, SVN_WC__CONFLICT_SRC_SUBVERSION))
> + {
> + *location = NULL;
> + return SVN_NO_ERROR;
> + }
> + c = c->next;
> +
> + repos_root_url = apr_pstrndup(scratch_pool, c->data, c->len);
apr_pstrmemdup() is more efficient. No need to look for NUL terminators.
> + c = c->next;
> +
> + if (c->is_atom)
> + repos_uuid = apr_pstrndup(scratch_pool, c->data, c->len);
> + else
> + repos_uuid = NULL;
> + c = c->next;
> +
> + repos_relpath = apr_pstrndup(scratch_pool, c->data, c->len);
> + c = c->next;
> +
> + SVN_ERR(svn_skel__parse_int(&v, c, scratch_pool));
> + revision = (svn_revnum_t)v;
> + c = c->next;
> +
> + kind_str = apr_pstrndup(scratch_pool, c->data, c->len);
> + node_kind = svn_node_kind_from_word(kind_str);
It would be nice to have an extern const svn_token_map for svn_kind_t
(and maybe for the old svn_node_kind_t), so that you can use
svn_token__from_mem(map, c->data, c->len) here.
> +
> + *location = svn_wc_conflict_version_create2(repos_root_url,
> + repos_uuid,
> + repos_relpath,
> + revision,
> + node_kind,
> + result_pool);
> + return SVN_NO_ERROR;
> +}
> +
> /* Get the operation part of CONFLICT_SKELL or NULL if no operation is set
> at this time */
> static svn_error_t *
> @@ -535,6 +588,7 @@ svn_wc__conflict_read_info(svn_wc_operat
> apr_pool_t *scratch_pool)
> {
> svn_skel_t *op;
> + svn_skel_t *c;
const
>
> SVN_ERR(conflict__get_operation(&op, conflict_skel));
>
> @@ -542,18 +596,35 @@ svn_wc__conflict_read_info(svn_wc_operat
> return svn_error_create(SVN_ERR_INCOMPLETE_DATA, NULL,
> _("Not a completed conflict skel"));
>
> + c = op->children;
> if (operation)
> {
> - svn_skel_t *what = op->children;
> -
> - int value = svn_token__from_mem(operation_map, what->data, what->len);
> + int value = svn_token__from_mem(operation_map, c->data, c->len);
>
> if (value != SVN_TOKEN_UNKNOWN)
> *operation = value;
> else
> *operation = svn_wc_operation_none;
> }
> - if (locations)
> + c = c->next;
> +
> + if (locations && c->children)
> + {
> + svn_skel_t *loc_skel;
> + svn_wc_conflict_version_t *loc;
const and const
> + apr_array_header_t *locs = apr_array_make(result_pool, 2, sizeof(loc));
> +
> + for (loc_skel = c->children; loc_skel; loc_skel = loc_skel->next)
> + {
> + SVN_ERR(conflict__read_location(&loc, loc_skel, result_pool,
> + scratch_pool));
> +
> + APR_ARRAY_PUSH(locs, svn_wc_conflict_version_t *) = loc;
> + }
> +
> + *locations = locs;
> + }
> + else if (locations)
> *locations = NULL;
>
> return SVN_NO_ERROR;
>
> Modified: subversion/trunk/subversion/tests/libsvn_wc/conflict-data-test.c
> URL:
> http://svn.apache.org/viewvc/subversion/trunk/subversion/tests/libsvn_wc/conflict-data-test.c?rev=1355577&r1=1355576&r2=1355577&view=diff
> ==============================================================================
> --- subversion/trunk/subversion/tests/libsvn_wc/conflict-data-test.c
> (original)
> +++ subversion/trunk/subversion/tests/libsvn_wc/conflict-data-test.c Fri Jun
> 29 22:05:08 2012
> @@ -448,6 +448,21 @@ test_serialize_text_conflict(const svn_t
> "theirs");
> }
>
> + {
> + svn_wc_operation_t operation;
> + const apr_array_header_t *locs;
> + SVN_ERR(svn_wc__conflict_read_info(&operation,
Could you please remember to keep a blank line between declarations
and code? It is hard to distinguish between them when run together
like this.
>...
Cheers,
-g
