Date:	Sun, 19 Oct 1997 17:19:03 +0200 (IST)
From:	Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To:	Bill Currie <billc AT blackmagic DOT tait DOT co DOT nz>
cc:	Paul Derbyshire <ao950 AT FreeNet DOT Carleton DOT CA>, djgpp AT delorie DOT com
Subject:	Re: WIP bug
In-Reply-To:	<199710170259.PAA26513@teleng1.tait.co.nz gatekeeper.tait.co.nz>
Message-ID:	<Pine.SUN.3.91.971019171837.27221T-100000@is>
MIME-Version:	1.0

On Fri, 17 Oct 1997, Bill Currie wrote:

> > The most recent WIP on Shawn's page has a bug of some kind. It makes
> > and runs perfectly. But, it displays a cryptic message about
> > something.html having a modification time in the future and a
> > warning that the build may be "incomplete", which it appears in fact
> > not to be.
> 
> Actully, it's a mis-feature in dos and (possibly) make: dos file 
> times only have 2 second granularity and make doesn't cope too well 
> with this on fast machines.  I get this all the time.

I wish that people would tell about such problems when they see them,
instead of silently dealing with them!  Bill, can you please send me a
simple Makefile that shows this bug, and tell me something about your
system hardware (e.g., how ``fast'' is it) and software?

During beta testing of Make 3.76, I heard reports about such problems
from a guy who maintains the Win32 port of GNU Make.  He said he saw
this on VFAT drives, and sent me a Makefile to exhibit the problem,
but I couldn't reproduce it, neither on Windows 95 nor on DOS (with
P166 and a lot of memory in both cases).  So Make 3.76.1 has the patch
that fixes this only enabled for Win32 version.  If Bill have told me
that he ``gets this all the time'', I would have asked him to see
whether that Win32 patch solves the problem, and we all would have had
a better Make now.

I also don't understand how this could happen, even on a very fast
machine.  AFAIK, DOS doesn't *round* the time to the nearest multiple
of 2 seconds: it *truncates* them.  So no file could ever have
modification time in the future, even if 2-sec granularity isn't taken
into account.  Maybe some DOS clones (OpenDOS?) do round, and then the
problem surfaces.  

But I need to see the problem before I try to fix it.  I cannot fix
something I'm unable to reproduce.  Please make a point of reporting
any such anomaly.  There's no reason for DJGPP ports to issue bogus
warnings, even if other DOS ports of Unix software are known to have
problems of this ilk which are usually dismissed as ``non-fatal''.

- Raw text -