Re: [RFC/PATCH 1/4] builtin: add git-check-mailmap command

From: Eric Sunshine <sunshine@sunshineco.com>
Date: Jul 11, 2013, 05:50 UTC
Message-ID: <CAPig+cRxuqOvRZQ5WZmUWruvEQytYM+RduYdkkN4_f8UdcuQnQ@mail.gmail.com>
In-Reply-To: <CACsJy8C6_ycum9RfxgfB0Xa2U7Fk02NyED5VvH6o1pExkcV6vA@mail.gmail.com>

On Wed, Jul 10, 2013 at 10:31 PM, Duy Nguyen <pclouds@gmail.com> wrote:

Show 9 quoted lines

> On Thu, Jul 11, 2013 at 2:03 AM, Eric Sunshine <sunshine@sunshineco.com> wrote:
>> +static const struct option check_mailmap_options[] = {
>> +       OPT_BOOLEAN(0, "stdin", &use_stdin,
>> +                   N_("also read contacts from stdin")),
>> +       OPT_BOOLEAN('z', NULL, &null_out,
>> +                   N_("null-terminate output lines")),
>
> I think OPT_BOOLEAN is deprecated in favor of OPT_BOOL (or OPT_COUNTUP
> if you really want -z -z -z to mean differently than -z)

Thanks, I knew this and intended to change it to OPT_BOOL but forgot. (The OPT_BOOLEAN was inherited from check-attr.c and check-ignore.c which I used as templates for check-mailmap.c.)

Show 11 quoted lines

>> +       maybe_flush_or_die(stdout, "contact to stdout");
>
> On error this function will print
>
> write failure on 'contact to stdout'
>
> maybe maybe_flush_or_die(stdout, "write contact to stdout") or
> something? From i18n point of view, maybe_flush_or_die should not
> compose a sentence this way. Let the second argument be a complete
> sentence so that translators have more freedom. But that's a different
> issue.

Indeed, it's not ideal. I chose "contact to stdout" for consistency with other callers, not because of a fondness for it.

  % git grep maybe_flush_or_die
  builtin/blame.c: maybe_flush_or_die(stdout, "stdout");
  builtin/check-attr.c: maybe_flush_or_die(stdout, "attribute to stdout");
  builtin/check-attr.c: maybe_flush_or_die(stdout, "attribute to stdout");
  builtin/check-ignore.c: maybe_flush_or_die(stdout, "check-ignore to stdout");
  builtin/check-ignore.c: maybe_flush_or_die(stdout, "ignore to stdout");
  builtin/hash-object.c: maybe_flush_or_die(stdout, "hash to stdout");
  builtin/rev-list.c: maybe_flush_or_die(stdout, "stdout");
  log-tree.c: maybe_flush_or_die(stdout, "stdout");

They seem pretty evenly split between just "stdout" and "foo to stdout". (I actually prefer plain "stdout" and will happily change it to that.)

-- ES