Re: svn commit: rev 4559 - trunk/subversion/libsvn_subr

[cm...@collab.net]

philip@tigris.org writes:

 Author: philip
 Date: 2003-01-24 06:40:32 -0600 (Fri, 24 Jan 2003)
 New Revision: 4559
 
 Modified:
    trunk/subversion/libsvn_subr/path.c
 Log:
 Fix another memory problem tracked down with valgrind.
 
 * subversion/libsvn_subr/path.c (svn_path_uri_encode): Allocate space for
   sprintf null.

I was wondering the other day if we shouldn't just do a single
svn_stringbuf_ensure here.  Either always do one at the beginning:

   svn_stringbuf_ensure(retstr, strlen (path) * 3) + 1);

   (This is worst case -- each char is replaced by three (%XX), and
    then room for the NULL at the end)

Or, do it the first time we find a char that needs replacing (someone
check my math):

   svn_stringbuf_ensure(retstr, copied + strlen (path + copied) * 3) + 1);

Anyway, just an idea.

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

Re: svn commit: rev 4559 - trunk/subversion/libsvn_subr

[cm...@collab.net]

Greg Stein gstein@lyra.org writes:

 On Fri, Jan 24, 2003 at 09:08:16AM -0600, cmpilato@collab.net wrote:
 ...
  I was wondering the other day if we shouldn't just do a single
  svn_stringbuf_ensure here.  Either always do one at the beginning:
  
     svn_stringbuf_ensure(retstr, strlen (path) * 3) + 1);
  
     (This is worst case -- each char is replaced by three (%XX), and
      then room for the NULL at the end)
  
  Or, do it the first time we find a char that needs replacing (someone
  check my math):
  
     svn_stringbuf_ensure(retstr, copied + strlen (path + copied) * 3) + 1);
  
  Anyway, just an idea.
 
 In apr_xml_quote_string, I coded a two-pass algorithm. One pass to count
 the special chars, do the alloc, then a second pass to make the changes. The
 neat part is that if no changes are necessary, it just returns the input
 string without an alloc/copy.

We could do the same pretty easily.  The test for URI-safeness is easy
and fast, and this beats XML-encoding by just a hair on the size test
because it's a flat 1-char-to-3-chars expansion for each unsafe byte.

Wanna code it up?

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

Re: svn commit: rev 4559 - trunk/subversion/libsvn_subr

[Greg Stein gs...@lyra.org]

On Fri, Jan 24, 2003 at 09:08:16AM -0600, cmpilato@collab.net wrote:
...
 I was wondering the other day if we shouldn't just do a single
 svn_stringbuf_ensure here.  Either always do one at the beginning:
 
    svn_stringbuf_ensure(retstr, strlen (path) * 3) + 1);
 
    (This is worst case -- each char is replaced by three (%XX), and
     then room for the NULL at the end)
 
 Or, do it the first time we find a char that needs replacing (someone
 check my math):
 
    svn_stringbuf_ensure(retstr, copied + strlen (path + copied) * 3) + 1);
 
 Anyway, just an idea.

In apr_xml_quote_string, I coded a two-pass algorithm. One pass to count
the special chars, do the alloc, then a second pass to make the changes. The
neat part is that if no changes are necessary, it just returns the input
string without an alloc/copy.

Cheers,
-g

-- 
Greg Stein, http://www.lyra.org/

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