From: Junio C Hamano <gitster@pobox.com>
Date: Thu, 24 Sep 2026 16:34:47 GMT
Subject: Re: [PATCH v5 1/3] refs: allow callers to supply old OIDs for batch deletion
Message-ID: <xmqq8q4q4nh4.fsf@gitster.g>
In-Reply-To: <CAOLa=ZTWq6eiqCwUyUhCffTn1=f9pdAip7nsYJMnaPPUuccB8g@mail.gmail.com>

Karthik Nayak <karthik.188@gmail.com> writes:

> I'm also still of the opinion that this shouldn't be done. A zeroed out
> null_oid is usually a user bug, where they haven't initialized a `struct
> object_id` correctly or ignored the return code while reading a ref.

... unless they are using an element in an object_array and want to
selectively have object names to some but not all of the elements in
the array.  In such a use case, a pointer to a null_oid is just as
good a representation as a NULL pointer of "N/A" for a parameter to
a function like this one that takes an optional object name.  You
could force each such callers to notice the entry they are about to
call this function with has a null_oid and pass a NULL instead, but
why force the caller to do so when the callee is capable of doing so
centrally?