bug#37683: Eww doesn't render Arabic script correctly

[Nicholas Drozd <nicholasdrozd@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 3986 bytes --]

I discovered that the problem only occurs when text is NOT scaled --
if the text is made larger or smaller by any amount (with
`text-scale-increase` etc), the cursive characters are joined
correctly.

A few screenshots are attached (hopefully their names are self-explanatory).

----------------------------------------

Bug report info:

In GNU Emacs 27.0.50 (build 6, x86_64-pc-linux-gnu, GTK+ Version
3.22.30) of 2019-10-09 built on ...
Repository revision: 6fa1558ca56c20226821e440a70129ece7b808b6
Repository branch: master
Windowing system distributor 'The X.Org Foundation', version 11.0.11906000
System Description: Ubuntu 18.04.2 LTS

Configured features:
XPM JPEG TIFF GIF PNG RSVG SOUND GPM DBUS GSETTINGS GLIB NOTIFY INOTIFY
ACL LIBSELINUX GNUTLS LIBXML2 FREETYPE HARFBUZZ M17N_FLT LIBOTF XFT ZLIB
TOOLKIT_SCROLL_BARS GTK3 X11 XDBE XIM THREADS PDUMPER LCMS2 GMP

Important settings:
  value of $LANG: en_US.UTF-8
  value of $XMODIFIERS: @im=ibus
  locale-coding-system: utf-8-unix

----------------------------------------

Not sure if this is the right command, but:

$ dpkg -l | grep "\(harf\)\|\(libm17n\)"
ii  gir1.2-harfbuz 1.7.2-1ubunt amd64        OpenType text shaping engine (GOb
ii  libharfbuzz-de 1.7.2-1ubunt amd64        Development files for OpenType te
ii  libharfbuzz-go 1.7.2-1ubunt amd64        OpenType text shaping engine ICU
ii  libharfbuzz-ic 1.7.2-1ubunt amd64        OpenType text shaping engine ICU
ii  libharfbuzz0b: 1.7.2-1ubunt amd64        OpenType text shaping engine (sha
ii  libm17n-0:amd6 1.7.0-3build amd64        multilingual text processing libr
ii  libm17n-dev:am 1.7.0-3build amd64        multilingual text processing libr

----------------------------------------

On Thu, Oct 10, 2019 at 1:37 AM Eli Zaretskii <eliz@gnu.org> wrote:
>
> > From: Nicholas Drozd <nicholasdrozd@gmail.com>
> > Date: Wed, 9 Oct 2019 12:16:16 -0500
> >
> > Here is the Persian Wikipedia page for Emacs:
> > https://fa.wikipedia.org/wiki/%D8%A7%DB%8C%D9%85%DA%A9%D8%B3
> >
> > When I open it in Eww, the cursive Perso-Arabic characters are not
> > strung together properly. For example, the article mentions the
> > "editor war", جنگ ویرایشگرها. In Eww, this is rendered as جنگ
> > ویرایشگرها with wrong forms (initial vs middle vs final vs isolated).
> > `D` runs `eww-toggle-paragraph-direction`, but that doesn't have any
> > effect on the problem. I think this might be a cursive issue rather
> > than a directionality issue, but I could be wrong.
> >
> > (As I prepare to send this, I notice that Gmail won't render that
> > phrase incorrectly, so the issue might not be communicated clearly
> > here.)
>
> Thank you for your report.
>
> First, you cannot effectively report text-shaping (and in general,
> rendering) problems by pasting the problematic text into an email
> message, because most Emacs maintainers use Emacs to read email, so
> any problems you report will be invisible to them, especially if you
> show two similar segments of text that are supposed to be rendered
> differently.  The only reliable way of reporting such rendering
> problems is by including screenshots of the wrong and the correct
> rendering (the latter from some other application or from a different
> Emacs version, if that's applicable).
>
> And second, you didn't provide any details about the Emacs version you
> used and how it was built (i.e. the features it can support) -- this
> information is collected by "M-x report-emacs-bug" when you invoke
> it.  My first suspicion is that you are using Emacs 26 or older and/or
> an outdated version of the libm17n library.  Emacs 27 uses HarfBuzz by
> default (if available on the system), and I cannot imagine that
> HarfBuzz doesn't render Farsi text correctly.  (If you send
> screenshots, I can verify that on my system, as I don't read Farsi, so
> I need an image of correct display to compare to.)

[-- Attachment #2: firefox-good.png --]
[-- Type: image/png, Size: 198016 bytes --]

[-- Attachment #3: eww-big-good.png --]
[-- Type: image/png, Size: 143545 bytes --]

[-- Attachment #4: eww-small-good.png --]
[-- Type: image/png, Size: 127575 bytes --]

[-- Attachment #5: eww-default-bad.png --]
[-- Type: image/png, Size: 211209 bytes --]