Re: svn commit: r997203 - in /subversion/trunk: ./ subversion/include/ subversion/libsvn_client/ subversion/libsvn_diff/ subversion/libsvn_fs_fs/ subversion/libsvn_ra_svn/ subversion/libsvn_repos/ subversion/libsvn_subr/ subversion/libsvn_wc/ subversion/sv...

[Stefan Sperling <st...@elego.de>]

On Wed, Sep 15, 2010 at 06:52:07AM -0000, hwright@apache.org wrote:
> ==============================================================================
> --- subversion/trunk/subversion/include/svn_string.h (original)
> +++ subversion/trunk/subversion/include/svn_string.h Wed Sep 15 06:52:06 2010
> @@ -253,6 +253,15 @@ svn_stringbuf_chop(svn_stringbuf_t *str,
>  void
>  svn_stringbuf_fillchar(svn_stringbuf_t *str, unsigned char c);
>  
> +/** Append a single character @a byte onto @a targetstr.
> + *
> + * reallocs if necessary. @a targetstr is affected, nothing else is.
> + * @since New in 1.7.
> + */
> +void
> +svn_stringbuf_appendbyte(svn_stringbuf_t *targetstr,
> +                         char byte);
> +

The docstring should list advantages svn_stringbuf_appendbyte(buf, c)
has over svn_stringbuf_appendbytes(buf, &c, 1). We need to understand
where the performance benefits really come from. IIRC the benefit was
dependent on the optimizer in the compiler to some extent, is this correct?

Stefan