bug#31316: Emacs hangs in `font_open_entity'

[Eli Zaretskii]

> Date: Mon, 30 Apr 2018 22:20:13 +0200 (CEST)
> Cc: 31316@debbugs.gnu.org
> From: Werner LEMBERG <wl@gnu.org>
> 
> The font is called `emmentaler-brace.otf' (part of lilypond,
> containing system braces for musical scores).  However, I wonder how
> this font can ever be considered as a fallback, since its SFNT (3,1)
> Unicode cmap contains only PUA character codes in the range
> 0xe100-0xe33f (and `fc-list -v' correctly lists that).  In other
> words, this font definitely doesn't contain anything relevant to the
> CJK character codes originally reported.

Please show a C-level backtrace from a breakpoint in that loop.  And
if the breakpoint breaks more than once when you do nothing after
invoking Emacs as shown in your OP, please show the backtraces from
all the times that breakpoint breaks.

Maybe looking at the backtrace will help us understand why Emacs
attempts to open that font.  Two general remarks I can make at this
point are that (a) Emacs decides whether a font might support a
character _without_ opening it (because opening a font is expensive,
and doing that for hundreds of fonts on a typical system will make
Emacs very annoying), and (b) only some font back-ends use Fontconfig,
so the fact that fc knows something doesn't yet mean Emacs does.