Re: Confused about source code of coreutils

[Bob Proulx]

Eric Blake wrote:
> Or you could use the relatively new git concept of a shallow clone, which
> only grabs the recent state of the repository without all the history.
> You can't do as much with a shallow clone, but if all you wanted was a
> snapshot of the current state, it fits the bill for smaller disk usage.

Additionally (of course Eric knows this but carrying on the discussion
for others...) it is possible to share the repository across multiple
working copies.  So if someone is were wanting to create several
different working copies for several different things it is not
required to use another 54MB for every git clone working copy.

Assuming:

  git clone git://git.sv.gnu.org/coreutils

Then:

  git clone -l --reference $PWD/coreutils git://git.sv.gnu.org/coreutils 
cu-test1

That will reference the files already in the $PWD/coreutils/.git depot
without duplicating them on the disk.  This works because git depots
are effectively static with regard to the previous history.  The new
cu-test1 depot created above requires that the referenced depot exist
but saves on duplicating the disk space.

It is also useful to daisy chain depots forward.  If a full depot has
been cloned once then it may be used to clone the next local depot.
This is very fast on a local disk and the depot created does not
depend upon the cloned depot.

  git clone -l $PWD/coreutils cu-test2

That will hardlink the static files from the previous depot history
into the newly created depot.  The disk space is then shared between
them by the use of hard links.  du will report file space consumed in
both but because it is hard linked the file space is actually shared.

The advantage of the hardlinks is that the depot is independent of the
first depot.  Either may be removed and the other will be fully
intact.  The disadvantage is that in this last case the automatic
setup of the "origin" URL is to the depot cloned which is not the
upstream depot but is instead the local disk depot.  I guess it is not
fully independent because of this but that is the only dependency.
The origin URL would need manual editing to point it upstream.

Git takes a toolbox approach.  A lot of tools are available and the
choice of how to use them is yours.  But you don't need to use all of
the tools all at once.  Use only the ones that you wish and leave the
others for another time.

Bob