Re: [PATCH v2] Refactor recv_sideband()

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Jun 28, 2016, 16:21 UTC
Message-ID: <alpine.DEB.2.20.1606281820060.12947@virtualbox>
In-Reply-To: <xmqqr3bh9wjh.fsf@gitster.mtv.corp.google.com>

Hi Junio,

On Tue, 28 Jun 2016, Junio C Hamano wrote:

Show 15 quoted lines

> Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
> 
> > Erm, sorry...
> >
> > On Tue, 28 Jun 2016, Johannes Schindelin wrote:
> >
> >> [...] we actually do not override fprintf() at all anymore [*1*] [...]
> >
> > ... forgot the
> >
> > Footnote *1*: In Git for Windows, we actually *do* override fprintf(),
> > thanks to using gettext, but it is *gettext* that is doing the overriding
> > now, not Git.
> 
> That's nice to know but is not particularly actionable.

Right. It was only meant as a footnote for interested parties. In Git for Windows, we do not need to override fprintf() to support color any longer.

> Do you mean
> 
>  - We do color correctly without having to override fprintf() these
>    days, so feel free and safe to use either fwrite() or write()

This one.

Show 8 quoted lines

>  - Because we do not override fprintf(), we no longer do colors
> 
>  - Or something in between, e.g. "we don't override fprintf()
>    ourselves but gettext does X for us (for unknown value of X), so
>    you will get color as long as you do Y (for unknown value of Y)"?
> 
> In other words, based on this piece of knowledge you shared with us,
> what is your recommendation for Lukas's patch to do?

Sorry for having been so obscure. I tried to clarify something that did not need clarification (it is of nobody's concern that fprintf() is overridden by gettext).

The short version is: do not worry about using fwrite().

Ciao, Dscho