Re: Using VC for change descriptions

[Paul Eggert]

On 01/09/2018 09:32 AM, Alfred M. Szmidt wrote:
> Here is an example where diffs simply do not work, when you modify the
> top of a function.  It will show the function that precedes it as the
> one being modified.

The diff works just fine if you read it in the obvious way, which is 
that it changes the API for frame-spec-choose by adding an argument 
no-match-retval. There's no other plausible way to read the diff. 
Obviously one should ignore the irrelevant part of the diff's context, 
and that includes the "face-set-match-display" as well as the "match))".

If this diff was intended to illustrate why ChangeLog entries are 
essential because we can't deduce things from diffs, then I'm afraid 
that it's an extremely weak illustration; it's actually evidence to the 
contrary.

> @@ -1512,13 +1512,15 @@ face-spec-set-match-display
>       match))
>   
>   
> -(defun face-spec-choose (spec &optional frame)
> -  "Choose the proper attributes for FRAME, out of SPEC.
> -If SPEC is nil, return nil."
> +(defun face-spec-choose (spec &optional frame no-match-retval)
> +  "Return the proper attributes for FRAME, out of SPEC.
> +
> +If no match is found or SPEC is nil, return nil, unless NO-MATCH-RETVAL
> +is given, in which case return its value instead."