pdfroff in groff 1.23.0.rc3 changes compared to 1.22.4

[G. Branden Robinson]

At 2023-04-04T13:48:39+0000, Michał Kruszewski wrote:
> Thanks for you response Branden.
> I use ms.

Aha, thanks!  I started trying to construct a reproducer using mm, then
noticed how terrible our description of the "TC" macro in the
groff_mm(7) man page is, so I've undertaken to improve it.

A bonus for the forthcoming groff 1.23.0.rc4, perhaps...

> Fortunately, the document with Makefile is available on github:
> https://github.com/Functional-Bus-Description-Language/Specification.
> You can easily compile the document by simply executing make.  Version
> compiled with 1.22.4 is even available in .pdf format in the
> repository.

At a quick look I think the problems involve the use of pdfroff.

If I try to format your document for PostScript using groff 1.22.4, I
get lots of problems as well.  Worse ones, in fact.  I see that the
header margin is reduced to zero, the cover page doesn't appear at all,
the typeface isn't restored in some places (it gets "stuck in Courier"),
and so on.

Part of this, I suspect, can be chalked up to pdfroff performing some
initialization of the ms(7) macro package behind the scenes, relieving
the user of that responsibility.

If I'm correct, then unfortunately this means that pdfroff may make an
ms(7) document seem valid when in fact it is not.  Your input looks
carefully composed, so I am hopeful that there is a simple fix, but I
tried some work-arounds, sprinkling calls of internal ms(7) "@init" and
"par@init" macros around in various places, and those didn't
significantly improve the situation.  (In one case I did get the header
margin back.  But I'd hate to recommend those as anything more than
emergency workarounds; calling ms(7) internals from a document is to be
discouraged.)

Furthermore, pdfroff author Keith Marshall is no longer maintaining it
as part of the groff distribution.  He has a new site for it
<https://osdn.net/users/keith/pf/groff‐pdfmark/wiki/FrontPage>.  It
could be that that version does not exhibit these problems, but I don't
follow its development, as groff itself keeps my hands full.

This was a fairly late-breaking change in groff 1.23.0's (very long)
release cycle, so we haven't decided yet if we want to simply withdraw
pdfroff (technically, the contrib/pdfmark directory in the sources) from
future versions of groff, or maintain our own version.  (We might have a
volunteer for the latter but I haven't gotten an update from them
lately.)  The plan had been to defer this decision until after 1.23.0's
final release, but if a lot of "pdfroff -mspdf"-using documents are
going to go haywire, we may need to make the decision sooner.  Or at
least warn pdfroff users of this possibility in the release
announcement.

> I have checked both PAGE=A4 and PAGE=letter and it looks like there is
> no difference.

Yes, I no longer think the page format has much to do with the problems
you're seeing.  (It might constitute a minor additional issue on top of
the much worse ones that are evident.)

I was about to start bisecting our repo so I could at least see where
pdfroff or groff ms changes started to make things go wrong, but
disappointingly, in spite of its claim for support of groff's "-a"
option in its man page, pdfroff does not in fact support it.[1]  I was
going to use that to try to speed the evaluation process at each stage
of the bisection.  Since there have been over 4,500 commits to the groff
Git repository since 1.22.4 was tagged, bisections can be lengthy.

I'll have to think a bit about the best way to attack this problem.

I'm open to suggestions from the mailing list.

Regards,
Branden

[1] observe: echo "hello, world" | pdfroff -a

signature.asc
Description: PGP signature