Re: Test failure: manpage writer

Hi Günter,

At 2023-11-04T19:11:19-0000, Guenter Milde via Docutils-develop wrote:
> thank you for the fast and comprehensive answer.

No worries; glad to be of help when I can.

> My patch adapts the "expected" output to what Docutils' manpage writer
> emits since the recent changes.
> 
> If I understand this right, then the changes work as intended, we can
> change the "expected" output sample and by this endorse the new
> manpage writer behaviour.

That sounds correct to me.  The change in groff 1.23.0's table
formatting for terminals was deliberate.  I didn't put this in the
"NEWS" file because there is no specification for tbl's behavior on
terminals with respect to whether box borders or vertical rules get
space "inboard" of them with respect to cell contents.

In fact, tbl(1) support on terminal devices is often neglected.
Documenter's Workbench troff and Heirloom Doctools troff produce fairly
dire output to terminals for anything but the simplest tables.

mandoc(1), by contrast, does fine (as does groff, of course).

> > Further, if you want a table where every cell is boxed, that's what
> > the "allbox" option is for.  You can use that instead of putting "|"
> > in your row descriptions and adding rows consisting solely of "_".
> 
> > tbl(1):
> >      allbox          Enclose each table entry in a box; implies box.
> 
> This would be even cleaner and the preferred solution for tables
> 
> * if it is supported by all relevant workflows

That, I cannot answer...

> * if support is not too new.

...this, I can.  The "allbox" feature is documented as far back as the
"tbl" white paper by Mike Lesk that was distributed with Seventh Edition
Unix.  That's 1979.  :)

(If you can find a copy of _Unix Time-Sharing System Programmer's Manual
Seventh Edition Volume 2A_ online, you'll find this region option
documented on or near page 155.)

Again, terminal output using DWB or Heirloom nroff is bad, but it's bad
either way, and has been for as long as the Python docutils have
existed.  My experiments show that "allbox" makes it neither better or
worse; I suspect the problem lies with computations around the line
drawing operations themselves.

Regards,
Branden

-----BEGIN PGP SIGNATURE-----
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=qE6c
-----END PGP SIGNATURE-----