Re: Forcing DOS line endings on checkout/update

[Laine Stump]

On Tue, Jul 31, 2001 at 02:43:38PM -0400, Alleman, Lowell wrote:
> > I have a hard time believing that no one else has needed this feature
> > before.
> > I'm surprised that there isn't some environmental variable that you
> > can just set, like 'CVS_ENV=DOS' which would setup the end of line modes.
> 

They have needed it. And every time they bring it up on this mailing
list, it is shot down by people who don't want/need it, and therefore
insist that it is a silly idea. I tired of the debate after the 2nd or
third time, and am now just waiting for Subversion to go gold - maybe
the people there will be more reasonable wrt real-world needs.

(So why didn't I just "do it and submit patches"? Why bother when I
know the patches will be rejected? And why am I replying to the thread
this time?  Because I figured it was time that the people asking for
this knew that they weren't crazy, and weren't alone).

As a matter of fact, WinCVS *does* implement this feature, but of
course that's going in the other direction from what you want.

Also, Philippe Payant submitted patches to the list back in Dec 2000,
to add this capability to CVS. If you want to try it and can't find
the message, I still have a copy.

Mike Castle <address@hidden> writes:

> CVS always has (and hopefully, always will!) used the default text mode of
> the system it is running on.
> 
> If you want it to check out things in DOS format, then run it on a machine
> that supports DOS format natively.

Sometimes that isn't practical, or isn't convenient, or you just plain
don't remember which platform you were running on when you did the
initial checkout (CVS informs you that you've done a subsequent
update/diff/commit/etc in different cryptic ways (ie "mysteriously
croaks with an unrelated message") depending on what the original
platform and the current platform are).

> I mean, what if you needed to share the files out to a MAC?  Would you
> expect cvs to check things using that file format as well?

Sure. So sue me.

> There are simply too many variables to get right.  Too much risk
> associated with getting it wrong.

The fact (or supposition, anyway) that something is difficult does not
mean that it shouldn't be done. It just means that care needs to be
taken when doing it. And anyway, somebody already *did* it.

> Your best solution is to set up some system on a win32 machine that, when
> a commit script is ran on the unix side, it sends some sort of message to
> the win32 (or Mac) machine.  When that message is received, it does a
> simple 'cvs up' and waits for the next signal.

Oh yeah, and *that's* not error prone? Anyway, Lowell's reason for
wanting this is just one from a long list of reasons why it would be
useful. Your "solution" would be little (if any) help in those other
situations.