Re: sg/commit-graph-cleanups (was Re: What's cooking in git.git (Jul 2020, #01; Mon, 6))

From: Taylor Blau <me@ttaylorr.com>
Date: Jul 7, 2020, 16:49 UTC
Message-ID: <20200707164953.GA36941@syl.lan>
In-Reply-To: <xmqqimez5l5q.fsf@gitster.c.googlers.com>

On Tue, Jul 07, 2020 at 09:45:21AM -0700, Junio C Hamano wrote:

Show 10 quoted lines

> Derrick Stolee <stolee@gmail.com> writes:
>
> >> Oops, I am not sure what happened here.  There I thought were issues
> >> pointed out at least on the latter series but was I hallucinating?
> >
> > I believe I have resolved all the open comments in both series.
>
> OK, then I do not have to be worried.
>
> Let's merge it down before -rc0.

Stolee and I did find a few unrelated bugs with Bloom filters in the past day or two while starting to use them at GitHub. The biggest is that Bloom filters are not used if the top-most incremental layer does not have Bloom data.

I'm not sure if this is a show-stopping bug or not, but if it is, I'd be happy to send a patch or two on top before -rc0.

The other bug is less serious, which is that we do reuse Bloom data from previous graphs even when writing a new graph with `--split=replace`. The fix is more involved, so I'm less inclined to rush it before -rc0.

Let me know your thoughts. Thanks.

> Thanks.

Taylor