On Mon, Feb 01, 2021 at 05:15:07PM +0000, Derrick Stolee via GitGitGadget wrote:
> From: Derrick Stolee <dstolee@xxxxxxxxxxxxx>
>
> Before checking if the repository has a commit-graph loaded, be sure
> to run prepare_commit_graph(). This is necessary because without this
> instance we would not initialize the topo_levels slab for each of the
> struct commit_graphs in the chain before we start to parse the
> commits. This leads to possibly recomputing the topological levels for
> commits in lower layers even when we are adding a small number of
> commits on top.
I think that the situation arises as follows:
- Prior to this patch, we didn't always prepare the commit-graph, so
the topo_levels slab wasn't initialized either.
- Then we try and compute the topo levels for new commits (which are
likely to be decendants of older ones that are in commit-graphs and
have their topo-levels already computed).
- But in the course of computing topo-levels for the new commits, we
have to recur on their ancestors, which *look* like they have
uncomputed topo levels.
That all makes sense, but it may be worth cutting and pasting what I
wrote into your patch to make it clearer.
> By properly initializing the topo_slab, we fix the previously broken
> case of a split commit graph where a base layer has the
> generation_data_overflow chunk.
Makes sense.
> -test_expect_failure 'lower layers have overflow chunk' '
> +test_expect_success 'lower layers have overflow chunk' '
> cd "$TRASH_DIRECTORY/full" &&
> UNIX_EPOCH_ZERO="@0 +0000" &&
> FUTURE_DATE="@2147483646 +0000" &&
Terrific :-).
Thanks,
Taylor
