bug#25493: 26.0.50; ediff merge should (optionally) show ancestor in fourth window

[Tino Calancha]

npostavs@users.sourceforge.net writes:

>> Subject: [PATCH 2/2] Show ancestor buffer in 3way merges
>>
>> Add an option ediff-show-ancestor', to control if the ancestor buffer
>> must be shown in 3way merges (Bug#25493);  set it non-nil by default.
>                                            ^^
> Extra space there.
Thanks.
>> -
>> -    ;; In case of merge job, fool it into thinking that it is just doing
>> -    ;; comparison
>> -    (let ((ediff-setup-diff-regions-function 
>> ediff-setup-diff-regions-function)
>> -      (ediff-3way-comparison-job ediff-3way-comparison-job)
>> -      (ediff-merge-job ediff-merge-job)
>> -      (ediff-merge-with-ancestor-job ediff-merge-with-ancestor-job)
>> -      (ediff-job-name ediff-job-name))
>> -      (if ediff-merge-job
>> -      (setq ediff-setup-diff-regions-function 'ediff-setup-diff-regions3
>> -            ediff-3way-comparison-job t
>> -            ediff-merge-job nil
>> -            ediff-merge-with-ancestor-job nil
>> -            ediff-job-name 'ediff-files3))
>> -      (funcall ediff-setup-diff-regions-function file-A file-B file-C))
>> -
>> +    (funcall ediff-setup-diff-regions-function file-A file-B
>> +             (if ediff-merge-with-ancestor-job file-Ancestor file-C))
>
> It works when I tested it, so I guess this is correct, but it's not
> really obvious to me why (i.e., why we no longer need to "fool it").
> Could you add an explanation to the commit message?
Now I mention that this trick is not necessary.  Before the patch there
were several parts of the code executed only if
`ediff-3way-comparison-job' was non-nil.  After the patch, those parts
of the code also test if `ediff-merge-with-ancestor-job' is non-nil;
that's why we don't need to fool it anymore.
> Otherwise looks good to me.
Thank you.
Pushed to master as commit 0f3d1b782353fd1fc0ab5f89d47d9e790f44e6b2