Re: isn't variance adjusted patching horribly dangerous?

       Philip Martin:

> Tom's example was contrived to give a conflict, I can
> contrive one that does not

You need not. Kfogel's document contains some examples. Mentioning
the existence of such examples is a logical non-sequitor.

Another example of where v.a.p. will silently and incorrectly supress
a desirable conflict would be the ordering of two function calls.

If the trunk has:

        foo (a, b);
        bar (x, y);

but the branch:

        bar (x, y);
        foo (a, b);

and the trunk changeset makes that `foo (a, b, c)', then v.a.p. can
silently create:

        foo (a, b, c);
        foo (a, b);

(By the "in-range edited line" rules.)

For that matter, with:

        foo (a, b);

in the trunk, and

        foo (munge(a), b);

in the branch, and a trunk change to:

        foo (a, b, c);

v.a.p. can silently give you:

        foo (a, b, c);

where

        foo (munge(a), b, c);

is what a human forced to resolve a conflict would want.

All that said: I'd surely like a stand-alone diff4 and even a built-in
diff4 (that emits a patch set rather than munging my tree) as a tool
to have around -- it's just a bloody mess, though to apply diff4
patches blindly as a day-to-day merging technique. My (possibly
faulty) understanding from off-list conversations was that blindly
applying diff4 was begining to be thought of as a way around the "back
and forth" merging problem.

-t

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Apr 10 02:09:49 2003