create()

Kenneth Brody kenbrody at spamcop.net
Wed Sep 4 11:24:19 PDT 2013


On 9/4/2013 11:48 AM, Mike Schwartz wrote:
>> I think this is one of those cases where, many years ago, someone thought
>> they needed to let the computer "catch up", and added a sleep (or
>> equivalent), and then (by sheer coincidence) things "magically" started to
>> work, and now they simply drop in a sleep whenever they have some
>> problem with external programs using filePro-created data.
>>
>> --
>> Kenneth Brody
>
>       You might be referring to somebody else, but, circa 1999, I did run
> into a repeatable scenario on a Windows 98 networked filePro 4.5 system.
> Because it turned out to be related to the high-powered graphics cards that
> were somehow delaying the data writes on the old IDE controllers, it took me
> several days to figure out.  The user needed the high-powered graphics cards
> to stay in the computers, so I added increasingly long sleep commands until
> I was sure the export files had all completed.

Even if the video card were to delay writes to the physical HD, there is no 
reason why anything not reading directly from the physical HD should see the 
"old" data.  (And no "normal" program should be reading from the physical 
HD.)  As soon as the write() call returns, all written data should be 
immediately available to any other program read()ing that file.

>       When I encountered other MS-DOS/Windows systems with similar problems,
> it was easier to add a second or two worth of sleep commands than it was to
> do a lot of hardware troubleshooting.

As long as you realize that this is, at best, a workaround to some bug 
elsewhere in the system, and all you are doing is maybe masking a symptom, 
and are willing to live with the consequences if the workaround doesn't work 
in all cases, so be it.  (As someone else noted, if the data isn't available 
immediately, who says that you are guaranteed that it is available after 
sleeping for some amount of time?)

-- 
Kenneth Brody


More information about the Filepro-list mailing list