Re: [RFC PATCH 2/3] usage: prefix all lines in `vreportf()`, not just the first

From: Junio C Hamano <gitster@pobox.com>
Date: May 28, 2018, 21:45 UTC
Message-ID: <xmqq8t83ihd7.fsf@gitster-ct.c.googlers.com>
In-Reply-To: <CACsJy8CgzDfbES2jC+Rv+bKj_JAk8dzEDCm89JMuBFQmyGg9Gw@mail.gmail.com>

Duy Nguyen <pclouds@gmail.com> writes:

Show 13 quoted lines

>>> +error:       sub/added
>>> +error:       sub/addedtoo
>>> +error: Please move or remove them before you switch branches.
>>>  Aborting
>>>  EOF
>>
>> This shows the typical effect of this series, which (I subjectively
>> think) gives us a more pleasant end-user experience.
>
> Also, very subjectively, I'm torn about this. To me, just one
> "error/warning/fatal" at the start of the first paragraph feels much
> better. If we have to somehow mark the second paragraph that "this is
> also part of the error message" then it's probably better to rephrase.

I personally can go either way. If you prefer less noisy route, we could change the function signature of vreportf() to take a prefix for the first line and another prefix for the remaining lines and pass that through down to the "split and print with prefix" helper.

That way, we can

 - allow callers to align 1st prefix (e.g. "error: ") with the
   leading indent for the second and subsequent lines by passing the
   second prefix with appropriate display width.

 - allow translators to grow or shrink number of lines a given
   message takes, and to decide where in the translated string to
   wrap lines.

Even though step 3/3 may become a bit awkward (the second prefix would most likely be only whitespace, and you'd need to write something silly like _("\t")), we can still keep the alignment if we wanted to.