[Info-vax] Cobol: Faster flushing of buffer to sequential file

[VAXman- at SendSpamHere.ORG]

In article <VdjsNP0JGKcS at eisner.encompasserve.org>, cornelius at eisner.decus.org (George Cornelius) writes:
>In article <iiicpm$gaf$1 at speranza.aioe.org>, "Richard Maher" <maher_rj at hotspamnotmail.com> writes:
>> I might have missed something in your post but am I correct in understanding 
>> it to have confirmed the fail case? (i.e. yes, no one else can see the 
>> record until it has been flushed or the file closed)
>
>Sorry, Richard, I guess I was a bit terse in my posts.
>
>Yes, your program as written, as Hein has remarked, does
>not do anything to update the EOF pointer in the file
>header.  Looks as if it takes SYS$FLUSH or a close to do
>that.  And apparently $ CONVERT/SHARE does not use strong
>enough voodoo (FAB$M_UPD in FAB$B_SHR) to get past that.
>
>But, I was taken by surprise when a simple
>$ OPEN/READ/WRITE/SHARE from DCL could not only access
>all data you had written, but, after a $ CLOSE
>operation, make it available to other readers as
>well.
>
>So a SYS$FLUSH or CLOSE is needed to cause the header
>to be updated, but an external program can do it.
>
>Thus, all three options for changing the app work -
>OPEN,WRITE,CLOSE; OPEN ALLOWING ALL; and SYS$FLUSH -
>but the second requires a reader program that uses
>the correct sharing semantics, or at least some
>DCL in the background that periodically opens for
>shared write and then immediately closes again.

And it requires the creator or initial opener to permit write access.
As I said somewhere in this long thread, I'm not one to allow writers
to my application log files.  Thus, the SYS$FLUSH seems to be the only
way to get the EOF update if the file is to remain open but allow read
access by others.

-- 
VAXman- A Bored Certified VMS Kernel Mode Hacker    VAXman(at)TMESIS(dot)ORG

All your spirit rack abuses, come to haunt you back by day.
All your Byzantine excuses, given time, given you away.