Re: There is a severe problem with EQN and square roots

"G. Branden Robinson" <[email protected]> Thu, 12 Feb 2026 07:23:56 -0600

Newsgroups	gmane.comp.printing.groff.general
Message-ID	<20260212132356.hcyt3ohbkvoruzur@illithid>

Hi Sigfrid,

Slightly editing my quote of you below
At 2026-02-12T11:17:01+0100, Sigfrid Lundberg wrote:
[...]
> I had actually forgotten that I had filed a bug report.
> 
> Inspired by the discussion I looked around in the tmac and font
> directories and found some odd definitions.

If you're referring to this:

$ git grep -nw sr tmac/ps.tmac
tmac/ps.tmac:30:.char \[radicalex] \h'-\w'\[sr]'u'\[radicalex]\h'\w'\[sr]'u'

...which is then fixed by...

> ...add[ing] another of my own (in my document)
> 
> .char \[sr] \[u203E]
> 
> then it seems as if the radical and the bar actually connect.

...then that sounds like the exact issue being tracked in Savannah
#63179...

https://savannah.gnu.org/bugs/?63179

...and further, it sounds like the font you're using does in fact _not_
define a radical extension glyph that is positioned, as I put it,
"weirdly as hell" relative to its bounding box.

> The square root symbol is too bold, but otherwise OK.

_That_ issue is being tracked as the one you filed, Savannah #67796.

http://savannah.gnu.org/bugs/?67796

I neglected to say what I thought could be done about that, beyond not
using \[sr] and \[radicalex] characters typeset at a larger size than
the radicand.  Merely "not using" them would typeset nothing in their
place, making radical signs invisible, a severe defect.

The alternative I had in mind but did not express was having eqn emit
(in troff mode) drawing commands to simulate the radical.

I'm attaching a crude sketch of what I have in mind.  Don't pay too
close attention to all the numerical figures in the *roff source; they
are present because I didn't use eqn at all to typeset this.

eqn aligns things much more nicely.  The thrust of the example is to
demonstrate the relative straightforwardness of drawing a radical sign
with `\D` escape sequences.  This can (and should, if chosen as the
solution to #67796) be refined to improve the appearance of the drawn
sign.  Improvements could include (a) using slightly different
thicknesses for the various strokes of the radical sign; (b) altering
the length of its leading horizontal stem; and (c) altering the angles
of the non-horizontal strokes.

Regards,
Branden

draw-radical.groff (text/plain, 450 B)

.\" Draw a big radical sign with a fraction radicand in GNU troff.
.sp 1i
Expression with \[lq]hand-drawn\[rq] radical sign:
.sp 2v
\Z"\
\D'l 1m 0'\
\D'l 1v 1v'\
\D'l 0 -3v'\
\D'l 4m 0'\
"\c
.\" Now move inside the radical sign.
\h'1m+1v+1n'\v'-1v'
.ft I
foobar
.br
\h'1m+1v+0.5n'\v'-1.5v'
\Z"\D'l 6n 0'"
.br
\h'1m+1v+1n'\v'-1.5v'
bazqux
.ft
.\" Do the same thing with GNU eqn.
.sp
Expression using
.ft I
eqn:
.br
.EQ
sqrt { foobar over bazqux }
.EN

draw-radical.ps (application/postscript, 7.2 KB) - not displayed

draw-radical.png (image/png, 27.3 KB) - not displayed

signature.asc (application/pgp-signature, 833 B)