original submission:

> The Texinfo manual says .ce centers text "without filling" ...
>
> On the other hand, the description of .ce in section 4.2 of
> CSTR #54 does not mention filling,

My second statement above is misleading: while strictly true that CSTR#54's .ce description does not mention "filling" by that name, it does say "A break occurs after each of the N input lines," which clearly communicates that no filling is done.  (Bjarni set me straight on this with his latest comment over in bug #46914.)

Thus the first statement above is groff's rewording of this longstanding principle.  As there's no controversy over any troff's filling behavior within a .ce block, this is an academic point regarding this bug, but I want to set the record straight, and/or scold my past self for the misreading.

No, I won't.  That's a different project's bug.

I mean bug #46914.  >:-|

Back in my court; updating Status and Assigned to.

> The Texinfo manual says .ce centers text "without filling" but is silent on breaking.

It seems groff's `ce` request implements a procedure that is used nowhere else in the formatter, such that lines are automatically broken when they reach the output line length but not filled.

This would look bad in conventional left-adjusted text (like reading emails in an MUA window too narrow for the text, and which wraps lines crudely), but presumably it's a sprinkle of DWIM behavior for .ce.

I'm inclined to consider this a documentation issue; to get authentic AT&T troff output, you need to explicitly turn off filling when using the `ce` request.

$ cat ATTIC/ce.roff
.ll 4i
.ce 100
This is a small text fragment that shows the differences
between the `.ce' (with and without `.nf`) and the `.ad c' requests.
.ce 0
.sp
.nf
.ce 100
This is a small text fragment that shows the differences
between the `.ce' (with and without `.nf`) and the `.ad c' requests.
.ce 0
.sp
.fi
.ad c
This is a small text fragment that shows the differences
between the `.ce' (with and without `.nf`) and the `.ad c' requests.
.pl \n(nlu
$ groff -Tascii ATTIC/ce.roff # 1.22.4, no difference in Git
This is a small text fragment that shows
            the differences
between  the  `.ce'  (with  and  without
    `.nf`) and the `.ad c' requests.

This is a small text fragment that shows the differences
between the `.ce' (with and without `.nf`) and the `.ad c' requests.

This is a small text fragment that shows
the differences between the `.ce' (with
 and without `.nf`) and the `.ad c' re-
                quests.
$ DWBHOME=~/dwb ~/dwb/bin/nroff ATTIC/ce.roff
This is a small text fragment that shows the differences
between the `.ce' (with and without `.nf`) and the `.ad c' requests.

This is a small text fragment that shows the differences
between the `.ce' (with and without `.nf`) and the `.ad c' requests.

This is a small text fragment that shows
the differences between the `.ce' (with
   and without `.nf`) and the `.ad c'
               requests.

That the first two specimens of these three produce distinct output in groff but not in DWB troff suggests to me that this behavior is deliberate.

By the way, you might have noticed that I tested with DWB troff rather than Heirloom.  This is because, to my knowledge, DWB 3.3 troff is the origin of Heirloom troff; the latter added many features.

You can try it out yourself.  It seems to have been modified only insofar as to make it buildable on modern Unix systems without too many screams from the compiler.

$ git remote -v
origin  https://github.com/n-t-roff/DWB3.3.git (fetch)
origin  https://github.com/n-t-roff/DWB3.3.git (push)

Tossing to you for feedback.