Windows native and system copy
Brian K. White
brian at aljex.com
Tue Mar 22 23:41:03 PST 2005
----- Original Message -----
From: "GCC Consulting" <gcc at optonline.net>
To: "'Kenneth Brody'" <kenbrody at bestweb.net>; "'Nancy Palmquist'"
<nlp at vss3.com>
Cc: "'Filepro 2 List'" <filepro-list at lists.celestial.com>
Sent: Tuesday, March 22, 2005 9:47 PM
Subject: RE: Windows native and system copy
>
>
>> -----Original Message-----
>> From: filepro-list-bounces at lists.celestial.com
>> [mailto:filepro-list-bounces at lists.celestial.com] On Behalf
>> Of Kenneth Brody
>> Sent: Tuesday, March 22, 2005 10:31 AM
>> To: Nancy Palmquist
>> Cc: Filepro 2 List
>> Subject: Re: Windows native and system copy
>>
>> Quoting Nancy Palmquist (Tue, 22 Mar 2005 10:15:56 -0500):
>> [...]
>> > Ken,
>> >
>> > While I understand you always work in a world where the OS works
>> > perfectly,
>>
>> Hardly. I've had my XP systems bluescreen, and I know what
>> horrors await if my 98 box goes to 0% resources available
>> (it's currently down to 15%). I even had a Unix box lock up
>> hard (no crash, no panic, just everything freezes for all
>> terminals immediately) when you tried to do the equivalent of
>> "unlink( (char *)-1L )" due to a bug in filePro at the time.
>>
>> > I also understand that we are talking about Windows and I
>> have yet to
>> > see any Windows version work perfectly. There are always driver
>> > issues, network issues, OS bugs and no matter how many patches you
>> > install it is still broken in one way or another.
>>
>> I've run into plenty of issues. However, I haven't (yet) run
>> into the problem of a file not being there until later.
>> Apparently, others have, and the SLEEP appears to work around
>> the system bug.
>>
>> > So if a small pause gives the wacky OS time to get itself together,
>> > then so be it.
>>
>> True. On the other hand, what's going to happen when the
>> system bug requires a 6-second delay to work around? ;-)
>>
>> --
>> KenBrody at BestWeb dot net spamtrap:
> Ken,
>
> Change the sleep to 6000 or something longer.
>
> All that aside, the file I am copying is there, the problem is, I think,
> that
> the OS still has it locked. It is "taking its own sweet time" to release
> it for
> other uses. :)
>
> Maybe this wouldn't be a problem if I ran the task from on of the XP
> workstations that are never shut
> off. Maybe the network latency would compensate for any problems with the
> task
> manager.
>
> Richard
Don't laugh.
A one second delay may actually BE the most technically correct answer
because it may actually be what windows itself does.
I just read a very interesting comment in a Linux magazine article in an
interview with a lead maintainer of Samba 3.x who works for HP.
Basically, the interviewer made a whimsical comment/prompt along the lines
of "bug for bug compatible with windows?" and the Samba developers response
was to describe just one example (presumably of many) of something they
found where their code initially performed exactly according to all
published specs. And yet, sometimes Excel will choke on a file shared via
Samba, claiming it was open by someone else when it really wasn't.
What they found was that Excel was opening the file, getting an oplock, and
then opening the file again right away without having released the first
lock, violating it's own lock. Samba was quite sanely reporting a sharing
violation on that second attempted open. Why does it only crash when the
file is hosted via Samba and not when hosted on a Windows share? Because
Windows does NOT generate a sharing violation for the 2nd open as long as
the first open closes within one second. He said that behaviour isn't
documented *anywhere* and they've also found differences between win2k and
2k3 that indicate that sambas regression testing procedures are apparently
more thorough than windows' own, else they'd have spotted the same
differences and either corrected or documented them.
The article is here:
LinuxUser, Issue 46, page 47.
The cover looks like this, it's probably still on some shelves, I just got
it today at a B&N, but it actually one issue old.
http://www.linuxuser.co.uk/index.php?option=com_content&task=view&id=58
Brian K. White -- brian at aljex.com -- http://www.aljex.com/bkw/
+++++[>+++[>+++++>+++++++<<-]<-]>>+.>.+++++.+++++++.-.[>+<---]>++.
filePro BBx Linux SCO Prosper/FACTS AutoCAD #callahans Satriani
More information about the Filepro-list
mailing list