The information needed here is a gathering of information about various *roff implementations.

This stuff looks like it's so ill-defined that I think we want to check out Seventh Edition Unix troff as well (which we can do with `tm` requests even if the program output is incomprehensible).

We should consider V7 troff, DWB 3.3, Heirloom Doctools, and possibly plan9port and neatroff.

Do the diversion measurement registers `dn` and `dl` measure:

• net motion of the drawing position,
• maximal motion of the drawing position relative to that where the diversion started,

or

• maximal bounds of spilled ink relative to the drawing position where the diversion started?

Negative motions seem ill-defined with only two registers.  Presumably this is why all `sp` requests mean `sp 1` inside diversions, but possibly this insight was insufficiently generalized back in the day.

Unassigning from self because I'm not the only person who can gather such information.

> Heirloom nroff reports the same zero width, so if this is indeed a bug, it's a widespread one.

If you consider the width of a diversion is the horizontal displacement of the longest line in the diversion, then the horizontal displacement in this single line example is zero, after outputting this diversion the horizontal movement is zero.

Using a diversion to capture the output of an eqn equation with the lineup indicator when there is an earlier equation with the mark indicator seemingly gives the wrong width of this lineup equation, in both troff and nroff.

In the example below, two different equations are processed: (1) with a mark indicator, and (2), printed once without and once with a lineup indicator, that is longer and has a common start with equation (1) before the indicators, such that output of both variants of (2) should look the same. However, the width of (2) with lineup indicator as reported by troff and nroff is not equal to the width of (2) without the lineup indicator (and equals the width of (1) in case of troff). For reference, I added the output of the latest (development snapshot) of Heirloom doctools, with the logical output.

$ cat mark_lineup_diversion
.di xx
.EQ
x mark = 1 + 1
.EN
.br
.di
\*(xx
.tm Width of (diversed) 1: \n(dl

.di xx
.EQ
x = 1 - 1 + 1 - 1 + 2
.EN
.br
.di
\*(xx
.tm Width of (diversed) 2: \n(dl

.di xx
.EQ
x lineup = 1 - 1 + 1 - 1 + 2
.EN
.br
.di
\*(xx
.tm Width of (diversed) 2 (using lineup): \n(dl

$ /usr/local/bin/neqn --version
GNU eqn (groff) version 1.23.0
$ /usr/local/bin/nroff --version
GNU nroff (groff) version 1.23.0
GNU groff version 1.23.0
Copyright (C) 2022 Free Software Foundation, Inc.
GNU groff comes with ABSOLUTELY NO WARRANTY.
You may redistribute copies of groff and its subprograms
under the terms of the GNU General Public License.
For more information about these matters, see the file
named COPYING.

called subprograms:

GNU grotty (groff) version 1.23.0
GNU troff (groff) version 1.23.0

$ cat mark_lineup_diversion | /usr/local/bin/neqn -Tascii | /usr/local/bin/nroff -Tascii | sed '/^$/d'
Width of (diversed) 1: 120
Width of (diversed) 2: 264
Width of (diversed) 2 (using lineup): 96
x=1+1
x=1-1+1-1+2
x=1-1+1-1+2
$ cat mark_lineup_diversion | /usr/local/ucb/neqn | /usr/local/ucb/nroff | sed '/^$/d'
Width of (diversed) 1: 120
x=1+1
Width of (diversed) 2: 264
x=1-1+1-1+2
Width of (diversed) 2 (using lineup): 264
x=1-1+1-1+2

$ cat mark_lineup_diversion | /usr/local/bin/eqn | /usr/local/bin/troff > /dev/null
Width of (diversed) 1: 36720
Width of (diversed) 2: 81390
Width of (diversed) 2 (using lineup): 36720
$ cat mark_lineup_diversion | /usr/local/ucb/eqn | /usr/local/ucb/troff > /dev/null
Width of (diversed) 1: 27086
Width of (diversed) 2: 58556
Width of (diversed) 2 (using lineup): 58556