new history punching code

Chris Mason chris.mason at oracle.com
Wed Nov 15 15:08:23 CST 2006


On Sun, Nov 12, 2006 at 08:10:15PM +0100, Ollivier Robert wrote:
> According to Chris Mason:
> > I've revived my history punching patch, you can pull it from here:
> > 
> > http://oss.oracle.com/mercurial/mason/punch
> 
> I just tried to merge with the curent tip and I had to manually merge 3/4
> files.  Do you think you could maintain it with respect to tip?  I'm not
> sure of what I choose when merging with kdiff3 so I may have made a
> mistake.

Testing and merging the punching code is non-trivial, so until we hash
out exactly how we want it to work, I don't plan on updating it
frequently.

My main concern with the punching code is that it adds complexity just
to save space.  It doesn't reduce the working set for O(N) operations,
and I think we may need a better partial history type patch to
accomplish that.

> 
> Context is, after a lenghty discussion with fellow FreeBSD committers at
> the EuroBSDCon summit just before the conference, the need to have a
> "obliterrate" kind-of feature just got higher in the FreeBSD priority
> list...
> 
> We have had previous requests from lawyers (the *etris and b*ggle cases)
> to remove actual code the source tree so even though it is much more
> difficult to have such a feature in modern VCS using hashes for everything
> and linking previous changesets data.

Since the punching code leaves the changeset in place, and may have to
keep certain revisions where a punch was requested to satisfy push/pull
requirements, I doubt it is a valid tool to resolve legal issues.

-chris



More information about the Mercurial mailing list