Re: usage message for newer-mtime in tar 1.13

[Mark Moraes]

Ok, thanks.  My question was less a need for
tar on windows (where I use cygwin), but because
I was adding a --older-mtime to do some replication
between some FreeBSD boxes here (large numbers of
new files, so tar is much better than rsync for this)
and I wanted to understand whether I should put my
code inside the ifdef !MSDOS or not.

Regards,
    Mark. 

----- Original Message ----- 
From: "Eli Zaretskii" <address@hidden>
To: <address@hidden>
Cc: <address@hidden>
Sent: Friday, December 07, 2001 12:56 AM
Subject: Re: usage message for newer-mtime in tar 1.13


> > From: "Mark Moraes" <address@hidden>
> > Date: Thu, 6 Dec 2001 18:47:10 -0800
> > 
> > 2. Should -N and its usage message really be conditional on !MSDOS?
> > All MSDOS that I'm familiar with have an mtime in their stat()
> > structures.
> 
> The fact that -N is disabled on MS-DOS is indeed a case of bad
> breakage, but please note that the current 1.13 series of GNU Tar
> don't support MS-DOS (and MS-Windows, for that matter) at all.  There
> are quite a few serious problems with some Tar features if the code is
> compiled unchanged on DOS/Windows machines, and -N is not the most
> serious of them.
> 
> As an example, consider the fact that many DOS/Windows filesystems
> record mtime with a 2-second granularity.  Figuring out the effects of
> this on -N, --diff, and similar Tar features is left as an exercise
> for the readers ;-)
> 
> A port of GNU Tar 1.12a to DOS/Windows was done in the past (by yours
> truly), but the changes are not yet rolled back into 1.13, since 1.13
> is based on sources that were forked before 1.12a was produced by the
> previous maintainer of GNU Tar.  If you are interested, or are in the
> need of a Posix-compatible Tar port for DOS/Windows, you can find it
> here:
> 
>   ftp://ftp.simtel.net/pub/simtelnet/gnu/djgpp/v2gnu/tar112ab.zip
> 
> See the file README.dos there for a list of problems in the original
> code that this port solves.
>