You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Ben Collins-Sussman <su...@collab.net> on 2005/08/09 16:28:57 UTC

question about r15253

I see that r15253 is nominated in the 1.2.2 STATUS file:


Index: subversion/libsvn_ra_local/ra_plugin.c
===================================================================
--- subversion/libsvn_ra_local/ra_plugin.c      (revision 15252)
+++ subversion/libsvn_ra_local/ra_plugin.c      (revision 15253)
@@ -920,6 +920,7 @@
           a loop.  Truly, Nothing Can Go Wrong :-).  But RA layers that
           go over a network should confirm the checksum. */
        SVN_ERR (svn_stream_copy (contents, stream, pool));
+      SVN_ERR (svn_stream_close (contents));
      }
    /* Handle props if requested. */


So the bug here is that ra_local's implementation of svn_ra_get_file 
() was calling svn_fs_file_contents(), reading all the data from the  
stream object, then forgetting to close the stream object.  This made  
perl bindings really unhappy.

I'm wondering if similar problems exist in other RA layers.  Looking  
in svnserve's call to svn_fs_file_contents(), it seems to close the  
stream after its done reading.  But looking in mod_dav_svn/ 
repos.c:dav_svn_deliver(), I don't see the stream being closed.  Is  
this a bad thing?


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: question about r15253

Posted by Branko Čibej <br...@xbc.nu>.
Ben Collins-Sussman wrote:

> I see that r15253 is nominated in the 1.2.2 STATUS file:
>
>
> Index: subversion/libsvn_ra_local/ra_plugin.c
> ===================================================================
> --- subversion/libsvn_ra_local/ra_plugin.c      (revision 15252)
> +++ subversion/libsvn_ra_local/ra_plugin.c      (revision 15253)
> @@ -920,6 +920,7 @@
>           a loop.  Truly, Nothing Can Go Wrong :-).  But RA layers that
>           go over a network should confirm the checksum. */
>        SVN_ERR (svn_stream_copy (contents, stream, pool));
> +      SVN_ERR (svn_stream_close (contents));
>      }
>    /* Handle props if requested. */
>
>
> So the bug here is that ra_local's implementation of svn_ra_get_file 
> () was calling svn_fs_file_contents(), reading all the data from the  
> stream object, then forgetting to close the stream object.  This made  
> perl bindings really unhappy.
>
> I'm wondering if similar problems exist in other RA layers.  Looking  
> in svnserve's call to svn_fs_file_contents(), it seems to close the  
> stream after its done reading.  But looking in mod_dav_svn/ 
> repos.c:dav_svn_deliver(), I don't see the stream being closed.  Is  
> this a bad thing?

It's a not exactly a resource leak since the stream will get closed when 
its pool is destroyed, but it's not nice, either.

I'm guessing that the bit with the Perl bindings and ra_local was more 
of a problem because there's no guarantee of regular pool cleanup.

-- Brane


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org