Re: Novel use of .char

[G. Branden Robinson]

Hi Deri,

At 2024-12-15T22:58:03+0000, Deri wrote:
> On Sunday, 15 December 2024 20:03:50 GMT G. Branden Robinson wrote:
> > I'm not able to reproduce this problem in my working copy--the GNU
> > head shows up, if fuzzily--and I haven't frotzed around in the
> > formatter since I last pushed.
> 
> Fuzzilly is to be expected - it is only 25x22 pixels.
> 
> I thought Peter's question was why did he need to add a kludge (a
> white period) to make it work? I confirmed the kludge was not required
> using 1.23.0.  Unfortunately, your test script (which works just as
> well as Peter's original) does not show the problem because it
> includes the kludge!!!

Yes, that was a screw-up on my part.  Thanks for steering me back on
track.

I can reproduce the problem, thanks to your attachments.

Attaching a screenshot of the salient part of the 1-page document.

I do now have an idea where I might have caused the regression.  It
might have happened when I was messing about in node.cpp on about
September 8.  Remember when you and I were talking about a certain
change making the page turn black when I used sboxes?  And all my
moaning about the selected font and drawing position, etc., and
whether device extension commands should be altering them, and my
piteous griping about the nonexistent documentation of this stuff?

Well, now I've got a reproducer so I really can glove up for the bisect,
and see how my intuition matches reality.

Crudely, it _looks_ like the baseline might be getting shifted upward by
the vertical dimension of the image, as if placement of the image-based
character "knew" it needed to compensate for the implicit vertical
motion caused by inlining it.  If so, I will be relieved, because we can
now regard that fixup as a kludge, and take it out.  If I can find it.

But I could be wrong, and there's no policy around any of this, just a
bunch of things that happened to work together, never got specced, and
never got unit-tested.  :(

> I have attached a version (sans kludge) which you can use for your
> testing, along with a pdf showing what output without the kludge looks
> like.

Thank you.  That's very helpful for times when my brain is switched off.
(I was in a hurry earlier today, preparing the house for company.
"Multitasking" gets me every time.)

> Please note I have changed the png to a pdf, since 1.23.0 only
> accepted images as pdfs, so it means you can use the same code for
> both groff versions.

That's helpful too.   The initial stages of the bisect will carry me
back in time long before your recent PerlMagick et al. support changes.

> PS You may notice I am one commit in front of you, this is because I
> get an error running bootstrap (after distclean) unless I add & commit
> gnulib beforehand (I then git reset HEAD~ after bootstrap because its
> not my code so I don't want to include it the next time I push origin
> master). Is there any way I can avoid this rigmarole?

I can reproduce this too if I "git pull" on my tablet.  I had made a
commit inside our gnulib submodule, attempting to sort of cherry-pick
(you can't _really_ cherry-pick because you're not on a proper branch)
Bruno Haible's fix for the math.h header in GCC 5.5 or so (the version
on Solaris 10).[1]

But I seem to have done it wrong--or it can't be done, because gnulib is
an external project.  I'll revert that, and when I have time, I'll ask
the gnulib folks if there is a right way to cherry-pick in a submodule.
I'm done with Solaris for the time being, and maybe by the time I get
back to it, there'll be a new gnulib tag that incorporates the fix.
(It's been 6 months since the last one; they're about due.)

The submodule problem didn't show up for me _on_ Solaris because I test
it with a distribution archive that I scp(1) over, not by using Git.

Sorry for the grief.  I was in a murderous fit, trying to kill *all*
Solaris compilation and testing issues at once because it is a major
pain to work on that platform and I always want to escape from it.

Regards,
Branden

[1] https://lists.gnu.org/archive/html/groff/2024-12/msg00078.html

problem-gnu-head.png
Description: PNG image

signature.asc
Description: PGP signature