"Johannes Schindelin via GitGitGadget" <gitgitgadget@xxxxxxxxx>
writes:
> + else if (!line.buf[0] || starts_with(line.buf, "index "))
> + /*
> + * A completely blank (not ' \n', which is context)
> + * line is not valid in a diff. We skip it
I noticed this while wondering how somebody could teach range-diff
to honor --notes=amlog while preparing the patches to be compared
[*1*], but this assumption goes against what POSIX.1 says these
days.
It is implementation-defined whether an empty unaffected line is
written as an empty line or a line containing a single <space> character.
cf. http://pubs.opengroup.org/onlinepubs/9699919799/utilities/diff.html#tag_20_34_10_07
We need to insert ", as we disable user's diff.suppressBlankEmpty
settings" before ". We skip it" (and if we get affected by the
setting, we need to fix it; it is not ultra-urgent, though).
[Footnote]
*1* ... which I do not have a good answer to, yet. As discussed
earlier, the diffopt passed into the show_range_diff() machinery is
primarily meant for the final output (i.e. how the matching patches
from the two iterations are compared) and not about how the patches
to be compared are generated. Worse, --notes=amlog (and possibly
other useful options) are parsed by "git log" side of the machinery,
not "git diff" side that populates diffopt.
