Re: NEW: cyrus-imapd-2.1.12

[Chris Giard <cgiard_(_dot_)_ports_(_at_)_magerealm_(_dot_)_com>]

> I found another issue which causes for a bit more concern:
>
> configure: warning: *** This system does not have a working mmap()
> configure: warning: *** Expect a considerable performance penalty
>
> Does OpenBSD really not have a working mmap()? :) As I understand this 
> happens because the corresponding conftest.c does not include unistd.h 
> and failes with an "implicit declaration of write()" and so on..

Based on what I have read on their discussion lists, the wording on 
this warning is not very accurate.  They are assuming that mmap and 
read/write can access a file at the same time, and changes from one 
will instantly show up in the other.  On a system with a unified memory 
model, this is true (Linux, FreeBSD, etc).  On OpenBSD, you have to 
explicitly call msync for the methods to see the same data.  The 
performance penalty they are talking about is their workaround for 
this.  I could be wrong on this, but that is what I have gathered from 
their lists.