Previously blinking was not exported in IE mode. IE still does not
support blinking (neither does Chrome or Safari) but the extra tag
does not make things any worse and allows importing the HTML back
to LibreOffice without loss of formatting.
Code is also simplified by removing conditionals for options that
no longer need to be disabled.
This disentangles another layer of intermediate substitution and assignment
which we can do now after Christina's work and aligns this a step closer to the
FcPattern concept and a few more percent more readable and maintainable I hope.
Rename ImplFontSelectData to FontSelectPattern because I can never remember the
name of the thing. This way I'll be able to remember that its basically the
equivalent of fontconfig's FcPattern
Same problem as a few commits earlier: libxslt's configure.js uses
debug=yes to mean using the debugging crt, which we don't want in just
a debug=t build. (And which will make libxslt.dll fail to load.) Look
at $(dbgutil), not $(debug). (And note that in general dbgutil builds
are disabled for Windows currently anyway.)
Passing the debug=yes option to libxml2's configure.js causes it to be
built against the debugging runtime, which is not what we want if we
just have used --enable-debug. It is with --enable-dbgutil that we
want to use the debugging runtime. So look at $(dbgutil), not
$(debug).
FreeSerif lacks glyphs in bold/italic variants that it has in the normal one. A
lot of our glyph fallback infrastructure, especially the caches don't expect
that a normal variant of a font with extra emboldening or extra font skew can
be a fallback for a bold/italic variant of itself which exists, but doesn't
have the missing glyphs that the normal one does.
We really need to improve our glyph/font caching, but we can get 90% of the
way there by excluding such cases from the caches.
Rename "shellnew" to "shlnew", and "2shellnew" then to "2shlnew" for
consistency.
See code in olenv/bin/modules/installer/windows/directory.pm:
create_unique_directorynames().
(Actually I think we don't even create any actual directory for
"shellnew" (and "2shellnew" even less), it is just a
placeholder... But oh well.)