The patch won't be merged as is.  Cases providing bad performance will be examined as they arise.

The full story behind this patch is that grep-2.5.1a does not handle UTF-8 gracefully at all.  The basic plan with handling UTF-8 in 2.5.1a is:

• whenever a buffer is parsed, go through the entire buffer deciding how many bytes make up each character
• use this information when necessary

This patch changes that to:

• when information about how many bytes make up a character is needed, work it out on demand

On the face of it, this is a small obvious improvement.  In fact it is much better than that, because the original scheme would calculate character lengths several times for each buffer: in fact, one full pass for every single potential match!

For a full discussion of this patch, as well as dfa-optional, including benchmarking results, see the mailing list.

This item is created to document in Savannah the existence of the following patch from Red Hat/Fedora:
  <http://cvs.fedora.redhat.com/viewcvs/devel/grep/>
  :pserver:anonymous@cvs.fedora.redhat.com:/cvs/dist
  grep-2.5.1-egf-speedup.patch

Here is the relevant log content:
  * Fri Jan  7 2005 Tim Waugh <twaugh@redhat.com>
  - Cache MB_CUR_MAX value in egf-speedup patch.
  * Fri Jan  7 2005 Tim Waugh <twaugh@redhat.com>
  - Fixed variable shadowing in egf-speedup patch.
  * Fri Dec 31 2004 Tim Waugh <twaugh@redhat.com>
  - Jakub Jelinek's much improved -Fi algorithm.
  * Tue Dec 21 2004 Tim Waugh <twaugh@redhat.com> 2.5.1-44
  - Fixed -Fi for multibyte input (bug #143079).
  * Thu Dec 16 2004 Tim Waugh <twaugh@redhat.com>
  - Bypass kwset matching when ignoring case and processing multibyte input (bug #143079).
  * Thu Dec 16 2004 Tim Waugh <twaugh@redhat.com>
  - Bypass kwset matching when ignoring case and processing multibyte input (bug #143079).
  * Tue Dec 14 2004 Tim Waugh <twaugh@redhat.com> 2.5.1-42
  - Further UTF-8 processing avoided since a '\n' byte is always an end-of-line character in that encoding.
  * Fri Dec  3 2004 Tim Waugh <twaugh@redhat.com> 2.5.1-41
  - Fixed a busy loop in the egf-speedup patch (bug #140781).
  Functional equivalence with pre-patched package.
  * Thu Nov  4 2004 Tim Waugh <twaugh@redhat.com> 2.5.1-34
  - More improvements to egf-speedup patch (bug #138076).
  * Thu Nov  4 2004 Tim Waugh <twaugh@redhat.com> 2.5.1-33
  - Small improvements to egf-speedup patch.
  * Wed Nov  3 2004 Tim Waugh <twaugh@redhat.com> 2.5.1-32
  - Remove mb-caching hack.
  - Better multibyte handling in EGexecute().
  - Don't need regex.c changes in grep-2.5-i18n.patch.