Web lists-archives.com

Re: [PATCH 2/4] revision walk: optionally use sparse reachability




On 3/12/2019 9:18 AM, Nathaniel Filardo wrote:
> The only caller that passes a non-zero value to prepare_revision_walk
> after this patch is builtin/pack-objects.  Without this, sparsity seems
> to do little good therein, as prepare_revision_walk will densely
> propagate UNINTERESTING flags from trees to tree contents, before
> mark_edges_uninteresting has a chance to be faster by being sparse.
> 
> Signed-off-by: Nathaniel Filardo <nwf20@xxxxxxxxxxxx>
> ---
>  bisect.c                         |  2 +-
>  blame.c                          |  2 +-
>  builtin/checkout.c               |  2 +-
>  builtin/commit.c                 |  2 +-
>  builtin/describe.c               |  2 +-
>  builtin/fast-export.c            |  2 +-
>  builtin/fmt-merge-msg.c          |  2 +-
>  builtin/log.c                    | 10 +++++-----
>  builtin/merge.c                  |  2 +-
>  builtin/pack-objects.c           |  4 ++--
>  builtin/rev-list.c               |  2 +-
>  builtin/shortlog.c               |  2 +-
>  bundle.c                         |  2 +-
>  http-push.c                      |  2 +-
>  merge-recursive.c                |  2 +-
>  pack-bitmap-write.c              |  2 +-
>  pack-bitmap.c                    |  4 ++--
>  reachable.c                      |  4 ++--
>  ref-filter.c                     |  2 +-
>  remote.c                         |  2 +-
>  revision.c                       | 10 ++++++----
>  revision.h                       |  2 +-
>  sequencer.c                      |  6 +++---
>  shallow.c                        |  2 +-
>  submodule.c                      |  4 ++--
>  t/helper/test-revision-walking.c |  2 +-
>  26 files changed, 41 insertions(+), 39 deletions(-)

This patch is very noisy. Perhaps the pattern established in 4f6d26b1:
"list-objects: consume sparse tree walk" should be reversed and instead
include a 'sparse_tree_walk' setting into 'struct rev_info'.

Changing so many method prototypes is rather invasive and unlikely to
benefit many of these callers.

If the setting is added to 'struct rev_info', then you'll want to remove
the parameter from mark_edges_uninteresting().

Thanks,
-Stolee