PATCH/RFC] allow delta data reuse even if base object is a preferred base

PATCH/RFC] allow delta data reuse even if base object is a preferred base

[Nicolas Pitre]

Signed-off-by: Nicolas Pitre <nico@cam.org>

---

Any reason why this couldn't be done?

When I reworked that code I preserved the existing logic wrt preferred 
base.  However I could not find a reason why it was that way.  Hence 
this patch.

diff --git a/builtin-pack-objects.c b/builtin-pack-objects.c
index 6db97b6..ee5f031 100644
--- a/builtin-pack-objects.c
+++ b/builtin-pack-objects.c
@@ -989,7 +988,7 @@ static void check_object(struct object_e
 		unuse_packed_git(p);
 		entry->in_pack_header_size = used;
 
-		if (base_entry && !base_entry->preferred_base) {
+		if (base_entry) {
 
 			/* Depth value does not matter - find_deltas()
 			 * will never consider reused delta as the

Re: PATCH/RFC] allow delta data reuse even if base object is a preferred base

[Junio C Hamano]

Nicolas Pitre <nico@cam.org> writes:

> Signed-off-by: Nicolas Pitre <nico@cam.org>
>
> ---
>
> Any reason why this couldn't be done?
>
> When I reworked that code I preserved the existing logic wrt preferred 
> base.  However I could not find a reason why it was that way.  Hence 
> this patch.
>
> diff --git a/builtin-pack-objects.c b/builtin-pack-objects.c
> index 6db97b6..ee5f031 100644
> --- a/builtin-pack-objects.c
> +++ b/builtin-pack-objects.c
> @@ -989,7 +988,7 @@ static void check_object(struct object_e
>  		unuse_packed_git(p);
>  		entry->in_pack_header_size = used;
>  
> -		if (base_entry && !base_entry->preferred_base) {
> +		if (base_entry) {
>  
>  			/* Depth value does not matter - find_deltas()
>  			 * will never consider reused delta as the

I was wondering about it myself the other day because I do not
remember the reasoning why I thought we do not want to reuse the
delta when the base object is not part of the resulting pack.
We would recompute and create the delta ourselves anyway if we
do not take this branch, so it is not about avoiding to delta
it.

I really do not see the reason now; I think it is safe and
reasonable to remove the check.