comment #5:

> In comment #3, the '-z' should have been '-Z'.

OK, now I can replicate your "groff" results.  I get identical -Z output whether I use groff 1.22.4 or a recent build.

nroff as shipped lacks the -Z option, so that one still fails.

> I now look at the space before the end of the output line to be a bug
> in the processing of the escape sequence '\:' in such a position.

Is this space perhaps due to the newline "echo" appends by default?  Consider the output of:

diff <(echo '\:' | groff -Z) <(echo -n '\:' | groff -Z)

(Notably, grops and gropdf produce identical output for both sets of intermediate output, so the "difference" here is effectively only internal.)

> The term for such "characters" (an element in a set) is "format
> effectors".

Then I amend my original bug report to say, ".char being unable to map something to a format effector, or at least to this particular format effector, is a bug either in the implementation, or in the lack of documentation of the restriction."

comment #3:

> "\:" is not a character but a movement in the scope of "groff".

I'm not sure what this means.  \: is not documented as a movement.

> echo '\:' | test-groff -z

Using either 1.22.4 or 1.23.0.rc1.2295-f55d, I get no output from that command (as expected, since on its own \: should produce no movement or output).  Does your test-groff script do something beyond the stock one?

> echo '\:' | test-nroff -z

groff includes no test-nroff script, so I know this is your own customization.  Thus, this result is only meaningful if you post your script.

comment #1:

> Alas, you didn't fix my typo.

Copy/paste failed to fix it.  I failed to notice it altogether.

(First mentioned as a side note in bug #58930, this problem is unrelated to that feature request and warrants its own report.)

.char \[u200B] \:

doesn't work.  .char being unable to map something to an escape, or at least to this particular escape, is a bug either in the implementation, or in the lack of documentation of the restriction.

Branden wrote the following in a follow-up comment:

It looks like the transform is just not happening:

.pl 1v
.char \[u200B] \:
.ds a \[u200B]
.length i \*a
\ni
8

.pl 1v
.ds a \[u200B]
.length i \*a
\ni
8

.pl 1v
.char a b
.ds a a
\*a
b

That unchanged length of 8, the exact character count of "\[u2000B]" is highly suspicious to me.