The following reply was made to PR os-freebsd/1310; it has been noted by GNATS.
From: Marc Slemko <[EMAIL PROTECTED]> To: =?KOI8-R?B?4c7E0sXKIP7F0s7P1w==?= <[EMAIL PROTECTED]> Cc: Apache bugs database <[EMAIL PROTECTED]> Subject: Re: os-freebsd/1310: FreeBSD have snprintf already long time but Apache still use its own Date: Thu, 23 Oct 1997 08:32:16 -0600 (MDT) On 23 Oct 1997, [KOI8-R] =E1=CE=C4=D2=C5=CA =FE=C5=D2=CE=CF=D7 wrote: > On Thu, 23 Oct 1997, Dmitry Khrustalev wrote: > =20 > > However, FreeBSD (BSTDIO) version is slower, resulting in ~1.5% > > performance degradation on zeusbench -k ( small file ). > =20 > Please specify FreeBSD version you test and where I can get zeusbench. It is originally from http://www.zeus.co.uk/ somewhere, but I'm not sure if it is there any more. http://www.worldgate.com/~marcs/fp/zb.c It takes some effort to conduct such tests. Testing of just the snprintf implementations showed a 5-10% difference between FreeBSD (2.2) and ap_snprintf. This may well be due to ap_snprintf being more specialized. It doesn't have all the features (eg. %q), but we can't use extra features anyway if they aren't portable. Unless there is a compelling reason (eg. ap_snprintf doesn't work) to use a system's built in snprintf, we really don't like to because it just introduces more hassles if it has bugs.
