From: ROBERT_FINLEY AT ntsc DOT navy DOT mil
Date: Mon, 04 May 98 17:55:59 EST
Message-Id: <9804048943.AA894329866@CCMAIL.NTSC.NAVY.MIL>
To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>, nate AT cartsys DOT com
Cc: djgpp AT delorie DOT com
Subject: RE: Make "Clock Skew" problem.
Precedence: bulk

More info:

I assume the changes to try were :

Change 1:

#ifdef WINDOWS32

becomes

#if defined (WINDOWS32) || defined (__DJGPP__)

and Change 2:

if (mtime > now && (((mtime %2) == 0) && ((mtime-1) > now)))

becomes

a.   if (mtime > now && (((mtime %2) == 0) && ((mtime-3) > now)))

or 

b.   if (mtime > now + 3)

I tested 6 permutations.

(Orig, Orig)            ->      doesn't work (Original Problem)
(Orig, Change 2a)       ->      doesn't work
(Orig, Change 2b)       ->      doesn't work

(Change 1, Orig)        ->      works
(Change 1, Change 2a)   ->      works
(Change 1, Change 2b)   ->      works

It turns out the Change 1 shuts-up make and Change 2 has no impact.   Now,
weather or not the messenger has been shot or the problem is fixed, I don't
know.


Thanks for your help,

Rob Finley



-----Original Message-----
From:   Eli Zaretskii <eliz AT is DOT elta DOT co DOT il> 
Sent:   Monday, May 04, 1998 6:10 AM
To:     nate AT cartsys DOT com
Cc:     ROBERT FINLEY; djgpp AT delorie DOT com
Subject:        Re: Make "Clock Skew" problem.


On Sun, 3 May 1998, Nate Eldredge wrote:

> !  if (mtime > now && (((mtime % 2) == 0) && ((mtime-3) > now)))

For a quick hack, I'd suggest to leave only the second part of the
test:
 if (mtime > now + 3)

But, if you think about this for a moment, you will realize that using
Make on a filesystem which lies up to 3 seconds about the file times
might totally disrupt the operation of Make.  Shutting up the warning
is tantamount to killing the messenger in this case: the problem
doesn't go away because you look the other way.

That is why I'm still thinking about the right way to solve this
problem.  Assuming all the files are on the same filesystem which has
the same time skew, I can solve this fairly easy.  But that assumption
will surely break in some cases, so Make will probably need to
maintain the skew value for each filesystem and apply it when checking
file times.

I also haven't given up hope that somebody will come up with a
reasonable explanation of why does this madness at all happen, and
that using that explanation we could find a better solution.