Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1a 12/4/83; site rlgvax.UUCP
Path: utzoo!linus!vaxine!wjh12!genrad!grkermit!masscomp!clyde!burl!ulysses!mhuxl!ihnp4!zehntel!hplabs!hao!seismo!rlgvax!guy
From: guy@rlgvax.UUCP
Newsgroups: net.unix-wizards
Subject: Re: Undocumented feature in Sys5: synchronous write
Message-ID: <1942@rlgvax.UUCP>
Date: Wed, 30-May-84 00:59:07 EDT
Article-I.D.: rlgvax.1942
Posted: Wed May 30 00:59:07 1984
Date-Received: Mon, 28-May-84 05:39:55 EDT
References: <267@pcsbst.UUCP>
Organization: CCI Office Systems Group, Reston, VA
Lines: 24

> Normally, writes to normal files are cached in the buffer pool and
> written out to disk with some delay.
> Some database people asked me to introduce a "write-through" system
> call, i.e. a write that writes data immediately to disk, thus avoiding
> disk consistency problems after a system crash. Looking at the Sys5 code,
> I found that the code for doing this is already there: in <sys/file.h>
> you have
> 	#define FSYNC   00020
> but an appropriate define
> 	#define O_SYNC  00020
> is missing from <fcntl.h> and also the docu doesn't mention it.

Yup.  Welcome to the wonderful world of UNIX; a treasure hunt each day.
The annoying thing is that some of the stuff are real treasures, but the
authors have seen fit to bury them deeply.  I suspect this feature was
introduced "accidentally"(!); it may have been a side-effect of the changes
to the file system code in System V to write file system data structures in
a specified order.  On the other hand, if somebody felt this wasn't useful
enough to applications to finish implementing (i.e., putting in the O_SYNC bit)
and documenting, I'd like to know what kind of applications they've ever seen.
At least Berkeley documented "fsync".

	Guy Harris
	{seismo,ihnp4,allegra}!rlgvax!guy