Re: [PATCH 1/2] write_entry(): fix misuses of `path` in error messages

From: Matheus Tavares Bernardino <matheus.bernardino@usp.br>
Date: Feb 9, 2021, 23:59 UTC
Message-ID: <CAHd-oW6CopkLyJTCErx9AYMw5YuhpUYhc7N4BH4sJKvy-7rvdA@mail.gmail.com>
In-Reply-To: <xmqqr1lpyls8.fsf@gitster.c.googlers.com>

On Tue, Feb 9, 2021 at 6:27 PM Junio C Hamano <gitster@pobox.com> wrote:

Show 12 quoted lines

>
> Matheus Tavares <matheus.bernardino@usp.br> writes:
>
> > The variables `path` and `ce->name`, at write_entry(), usually have the
> > same contents, but that's not the case when using a checkout prefix or
> > writing to a tempfile. (In fact, `path` will be either empty or dirty
> > when writing to a tempfile.) Therefore, these variables cannot be used
> > interchangeably. In this sense, fix wrong uses of `path` in error
> > messages where it should really be `ce->name`. (There doesn't seem to be
> > any misuse in the other way around.)
>
> Sounds reasonable.  Don't we want to protect this fix with tests?

Yeah, good idea. I will add a couple tests to check the error message on missing blobs and when trying to write a submodule to a tempfile. This should cover 3 out of the 4 error() calls changed in this patch. (The last one would be when ce->mode is unknown. I'm not sure if/how I can trigger that case.)