Re: [PATCH 0/6] clean up parsing of maybe_bool

[Stefan Beller <sbeller@google.com>]

On Mon, Aug 7, 2017 at 11:20 AM, Martin Ågren <martin.agren@gmail.com> wrote:
> When we want to parse a boolean config item without dying on error, we
> call git_config_maybe_bool() which takes two arguments: the value to be
> parsed (obviously) and a `name` which is completely ignored. Junio has
> suggested to drop `name` and rename the function [1]. That effort even
> started shortly after that, by introducing git_parse_maybe_bool(). The
> new function currently only has a single user outside config.c.
>
> Patch 5 of this series deprecates the old function and updates all
> callers to use git_parse_maybe_bool() instead. Patch 6 is a final
> cleanup: one of the converted callers suddenly had an unused argument.
>
> Patches 3 and 4 prepare for the switch. In particular, patch 4 ensures
> that the two functions are actually equivalent. In doing so, it affects
> `git push --signed=..` which was very slightly inconsistent to the rest
> of Git.
>
> Patch 2 adds a failing test in preparation for patch 4. Patch 1 corrects
> the documentation not to say "git push --sign=.." to make it a bit more
> obvious that the opposite typo is not being made in patches 2 and 4.
>
> git_parse_maybe_bool is used in sb/diff-color-move, which is in "next".
> This series makes --color-moved and diff.colormoved consistent with
> other booleans. That should be a good thing, but cc Stefan to be sure.

The series looks fine to me overall, though patch 5 is overly gentle IMHO.
We could have removed it right there as Junio is very good at resolving
conflicts or producing dirty merges for such a situation.
But delaying it until no other series' are in flight is fine with me, too.

Looking back at sb/diff-color-move and the code where
git_parse_maybe_bool is used, I do not think this will be an issue.

Thanks,
Stefan