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

[PATCH] svn_io_check_path, round 2 (was: Re: svn commit: r30931 - trunk/subversion/libsvn_client)

From: Daniel Shahaf <d.s_at_daniel.shahaf.co.il>
Date: Mon, 5 May 2008 10:45:16 +0300 (Jerusalem Daylight Time)

Daniel Shahaf wrote on Sat, 3 May 2008 at 00:19 +0300:
> kameshj_at_tigris.org wrote on Fri, 2 May 2008 at 06:05 -0700:
> > @@ -789,7 +789,8 @@ svn_client__make_local_parents(const cha
> > apr_pool_t *pool)
> > {
> > svn_error_t *err;
> > -
> > + svn_node_kind_t orig_kind;
> > + SVN_ERR(svn_io_check_path(path, &orig_kind, pool));
>
> According to svn_io_check_path's docstring (both before and after the
> recent change), when "intermediate directories on the way to @a path
> don't exist", it returns an error

I double checked; contrary to the docstring's promise, no error is
returned in this case. I try to fix the docstring in the patch below by
removing the distinction between "Path does not exist" and "Path's parent
does not exist" and redefining the "error" and "svn_node_unknown" cases to
match svn_node_kind_t's documentation.

I'll commit in a few days, but posting here for sanity check.

[[[
Follow up to r30937.

* subversion/include/svn_io.h
  (svn_io_check_path): Try again to match the docstring to the
    implementation.
]]]

The complete patched docstring is:

    /** Determine the @a kind of @a path. @a path should be UTF-8 encoded.
     *
     * If @a path is a file, set @a *kind to @c svn_node_file.
     *
     * If @a path is a directory, set @a *kind to @c svn_node_dir.
     *
     * If @a path does not exist, set @a *kind to @c svn_node_none.
     *
     * If @a path exists but is none of the above, set @a *kind to @c
     * svn_node_unknown.
     *
     * If unable to determine @a path's kind, return an error, with @a *kind's
     * value undefined.
     *
     * Use @a pool for temporary allocations.
     *
     * @see svn_node_kind_t
     */
    svn_error_t *svn_io_check_path(const char *path,
                                   svn_node_kind_t *kind,
                                   apr_pool_t *pool);

The diff is:

    Index: subversion/include/svn_io.h
    ===================================================================
    --- subversion/include/svn_io.h (revision 31020)
    +++ subversion/include/svn_io.h (working copy)
    @@ -79,15 +79,17 @@ typedef struct svn_io_dirent_t {
      *
      * If @a path is a directory, set @a *kind to @c svn_node_dir.
      *
    - * If @a path does not exist in its final component, set @a *kind to
    - * @c svn_node_none.
    + * If @a path does not exist, set @a *kind to @c svn_node_none.
      *
    - * If intermediate directories on the way to @a path don't exist, return
    - * an error, with @a *kind's value undefined.
    + * If @a path exists but is none of the above, set @a *kind to @c
    + * svn_node_unknown.
      *
    - * Otherwise, set @a *kind to @c svn_node_unknown.
    + * If unable to determine @a path's kind, return an error, with @a *kind's
    + * value undefined.
      *
      * Use @a pool for temporary allocations.
    + *
    + * @see svn_node_kind_t
      */
     svn_error_t *svn_io_check_path(const char *path,
                                    svn_node_kind_t *kind,

Daniel

> and leaves the value of orig_kind uninitialized.
>
> In svn_client__make_local_parents we cannot assume that the intermediate
> directories exist; therefore, the above call might raise an error.
> However, 'mkdir --parents' works in HEAD; an error is not raised.
>
>
> This leaves me with two questions:
>
> * Should we be calling svn_io_check_path in
> svn_client__make_local_parents ?
>
> * Is svn_io_check_path's docstring still wrong? (when it says errors
> would be returned)
>

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-05-05 09:45:49 CEST

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

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.