Re: [PATCH v2 2/4] bulk-checkin: remove global transaction state

From: Junio C Hamano <gitster@pobox.com>
Date: Aug 22, 2025, 20:25 UTC
Message-ID: <xmqqikifw1cl.fsf@gitster.g>
In-Reply-To: <njvcqkkcotesqznalieuwhzow7lxbejxlfw3knbhalmqjcb7cs@3vwea4r2zljl>

Justin Tobler <jltobler@gmail.com> writes:

> If it is preferable, I can combine these two patches together. I kept
> them separate as I thought it would better explain the steps, but maybe
> that isn't the best way to structure the patch as we are largely undoing
> the change in the next patch anyway.

This shouldn't make much difference, but I prefer the current "awkared but done as two steps" arrangement slightly better.

> I'm hoping for an eventual state where the transactional interfaces,
> {begin,end}_odb_transaction(), are moved directly into odb.{h,c}.

Yes.

Show 5 quoted lines

> The
> current implementation of transaction handling is specific to the
> current object database source. In a pluggable object database future
> where we could have different types of object database sources,
> transaction handling will likely have to implemented separately.

Yup.

> With this in mind, we could move the current transaction implementation
> into something like object-files.c. This is already where the vast
> majority of its call sites are and would enable us to further simply the
> interface we expose.

Great.