Knuth-Plass: adjacent demerits on the first line

[Karl Berry]

Didier (and all) - I asked DRF if he had any comments, and he did:

Date: Fri, 5 May 2023 14:36:22 +0000 (UTC)
From: David Fuchs

 I think Didier missed a detail: you get adj_demerits only when the
fitness changes by greater than 1. So, the first line of a paragraph
gets adj_demerits only when it's very_loose_fit; and that seems like a
reasonable design decision.  Oh, and especially when you consider that
the last line of the previous paragraph is extremely unlikely to
actually have been set as very_loose_fit (I suppose that will only when
you're doing some egregious full-justification on even the last line of
the paragraph.) Of course, I could be not understanding him correctly.

..or, as The TeXbook puts it (**emphasis** mine; perhaps the double use of
"adjacent" confuses things; and it's really subtle to note that the
"i.e." is not an "e.g." and means that all the cases are given, rather
than just example cases):

    Two adjacent lines are said to be {\sl {visually incompatible}\/} if
    their classifications **are not adjacent**, i.e., if a tight line is
    next to a loose or very loose line, or if a decent line is next to a
    very loose one.

and:

    If two consecutive lines are visually incompatible, in the sense
    explained a minute ago, the current value of\/ ^|\adjdemerits| is added
    to~$d$. 

-drf