Re: [PATCH] fetch-pack: avoid fixing thin packs when unnecessary

Hi,

On Mon, 18 Dec 2006, Nicolas Pitre wrote:

On Mon, 18 Dec 2006, Johannes Schindelin wrote:

quoted

When we know that there are no common commits, the pack must be
closed (i.e. non-thin) already. Avoid "fixing" it in that case.

Signed-off-by: Johannes Schindelin <redacted>

---

	On Mon, 18 Dec 2006, Johannes Schindelin wrote:

	> On Mon, 18 Dec 2006, Randal L. Schwartz wrote:
	> 
	> > But then it took nearly an *hour* at the next phase:
	> > 
	> >   Resolving 313037 deltas.
	> >    100% (313037/313037) done
	> 
	> Ouch.
	> 
	> We try to avoid unpacking the thin packs received by git-fetch.
	> This means completing that pack (since it can contain deltas 
	> against objects which are part of another pack).
	> 
	> However, for the clone this is utter overkill. We really should 
	> try to avoid resolving unnecessarily. This is really for the 
	> clone case, since we do not have _any_ objects in the local 
	> repository.
	> 
	> It happens that the other case -- fetching an independent branch 
	> -- is easy enough: we already have the check for it in 
	> fetch-pack.c:586.

	... and here is a lightly tested fix.

NAK.

This fixes nothing.  See previous message.

You're completely right. My patch does what I say, but it does not fix the 
problem.

Ciao,
Dscho

On Mon, 18 Dec 2006, Johannes Schindelin wrote:

Hi,

On Mon, 18 Dec 2006, Nicolas Pitre wrote:

quoted

On Mon, 18 Dec 2006, Johannes Schindelin wrote:

quoted

	... and here is a lightly tested fix.

NAK.

This fixes nothing.  See previous message.

You're completely right. My patch does what I say, but it does not fix the 
problem.

Your patch is also unnecessary.  For use_thin_pack to be true, you must 
provide --thin to git-fetch-pack which is not the case when cloning.