[ksh93-integration-discuss] [osol-code] [perf-discuss] extremly bad performance of Solarisregex

[Glenn Fowler]

On Thu, 01 Nov 2007 04:01:49 +0100 Roland Mainz wrote:
> ... since it's a multibyte locale. But for the GNU stuff there is a 3rd
> issue - since a while they add special UTF-8 hacks to speed the
> processing for such endings up (sometimes at the expense of being
> restricted to the BMP or similar things (e.g. if you have one charatcer
> outside the BMP your regex patter may not work anymore)) which means GNU
> tools must be tested with singlebyte locales, an UTF-8-based locale and
> a non-UTF-8-based locale... fun... ;-(

there's also a *big* hit for gnu sort w.r.t. UTF-8 LC_COLLATE

-- Glenn Fowler -- AT&T Research, Florham Park NJ --