Re: [GSoC][PATCH 1/1] sequencer: print an error message if append_todo_help() fails

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Jun 26, 2018, 21:52 UTC
Message-ID: <nycvar.QRO.7.76.6.1806262348340.21419@tvgsbejvaqbjf.bet>
In-Reply-To: <xmqqk1qljpe0.fsf@gitster-ct.c.googlers.com>

Hi Junio,

On Tue, 26 Jun 2018, Junio C Hamano wrote:

Show 30 quoted lines

> Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
> 
> > On Tue, 26 Jun 2018, Alban Gruin wrote:
> >
> >> This adds an error when append_todo_help() fails to write its message to
> >> the todo file.
> >> 
> >> Signed-off-by: Alban Gruin <alban.gruin@gmail.com>
> >
> > ACK.
> >
> > We *may* want to fold that into the commit that adds `append_todo_help()`.
> 
> Absolutely.  This looks more like an "oops, I made a mess and here
> is a fix on top", and even worse, it does not make an effort to help
> readers where the mess was made (iow, which commit it goes on to
> of); it is better to be squashed in.
> 
> I do not know offhand who Alban's mentors are, but one thing I think
> is a good thing for them to teach is how to better organize the
> changes with readers in mind.  The author of a patch series knows
> his or her patches and how they relate to each other a lot better
> than the readers of patches, who are reading not just his or her
> patches but the ones from a lot wider set of contributors.  Even
> though append-todo-help and edit-todo may have been developed as
> separate steps in author's mind, it is criminal to send them as if
> they are completely separate topics that can independently applied,
> especially when one depends on the other.  It is a lot more helpful
> to the readers if they were sent as a larger single series, because
> doing so _will_ tell the readers which order the dependency goes.

Chris & Stefan are Alban's mentors, and I spend quite a bit of my time on IRC, ready to help Alban when he has questions. Chris & Stephan mainly act as first-line reviewers.

> > And, as I mentioned previously, I would love for that function to be
> > used as an excuse to introduce the long-overdue `interactive-rebase.c`
> 
> I am not sure if I like this direction.

Blame me, not Alban. I am pretty familiar with sequencer.c, and I know that it is way too large.

> As newbies are often very bad at coming up with APIs and naming global
> functions, keeping everything as "static" inside a single sequencer.c
> tends to avoid contaminating the global namespace.

Then I just need to make sure to suggest good names that are safe for the global namespace, don't I?

Seeing as sequencer.c is so long already, it is own little mega namespace anyway, so we already have to be very careful *within* sequencer.c.

Ciao, Dscho