Re: [PATCH v2 06/23] builtin/pack-objects.c: mark more strings for translation

[Junio C Hamano <gitster@pobox.com>]

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

> @@ -149,7 +149,7 @@ static void *get_delta(struct object_entry *entry)
>  	delta_buf = diff_delta(base_buf, base_size,
>  			       buf, size, &delta_size, 0);
>  	if (!delta_buf || delta_size != DELTA_SIZE(entry))
> -		die("delta size changed");
> +		die(_("delta size changed"));

I wonder if this should be a BUG("..."), as it falls into the same
category as assert(), just like others turned into BUG("...") with
step 01/23.  If the following observation is correct, then we need
to turn this message into BUG() in 01/23 in a reroll and leave it
untranslated:

    In try_delta() we have earlier tried delta against many other
    objects and picked the best one as (base object, size of delta,
    delta chain length) tuple, and in the caller of this function,
    we write out the delta by recomputing it (because the resulting
    delta, unless it is trivially small, would have been discarded
    long before we come to this phase).  If diff_delta() between
    that chosen base object results in a different size from what we
    observed earlier, on which we based our decision to use this
    object as the base in the first place, we have a problem.