Am Freitag, 21. Juli 2017 um 21:30:35, schrieb Guenter Milde <mi...@users.sf.net> > On 2017-07-19, Kornel Benko wrote: > > > > > I did as you proposed (Dejavu font), but maybe you are unaware that > > With TL15 everything compiles here > > TL15 does not bind \textcompwordmark to the ZWNJ character but uses the > default definition also for Unicode fonts (EC1 and EC2). > > > With TL16 I get missing glyphs if font is Dejavu, but not with Free* > > Mark, that changing 'Dejavu Sans Mono' to e.g. 'Dejavu Sans' cures the > > compilation though > > With TL17 I get missing glyphs for any font. I have to replace the mono > > font to > > get it to compile. > > In TL16, the font encoding used by "fontspec" (Xe/LuaTeX with "system > fonts") changed to TU: > > * In this encoding, \textcompwordmark is mapped to the ZWNJ character. > > * Only some fonts actually contain this character. > > * LyX turns the "missing character" WARNING into an ERROR > This is usually a good thing (because printing characters that are missing > represent data loss). > > For ZWNJ, it would be better if the warning was *not* > turned into an error, because the ligature break works also if the > character is missing -- no data loss, correct output! > > > So, maybe better we could omit \textcompwordmark in mono fonts. > > This would only cure the special case tested in the export tests. It will > not help in documents with the default LM or custom fonts where the serif > or non-serif fonts miss the ZWNJ. > It would also not help in documents using the "ligature-break" special-char. >
But that is not our problem IMHO. If someone uses that glyph with a font not
having it,
it is not lyx's or latex problem.
But for mono-fonts, who may expect the glyph to exist?
WE output \textcompwordmark also for mono fonts, and this feels very wrong.
> The better cure is one of
>
> a) an exception list for the "turn missing character warning to error" code,
>
> b) preamble code reverting the binding of \textcompwordmark to ZWNJ (at
> least for the next years until "system fonts" can be safely expected to
> ship with the ZWNJ character).
>
> I prefer a), because the missing character still breaks the ligature (i.e.
> the output looks as expected).
Hm, yes, but that would also be a change in command line for lyx.
ATM, lyx is responsible to filter the latex log if instructed so.
# lyx --ignore-error-message missing_glyphs
With this method (ATM) _any_ missing glyph is ignored unfortunately.
> Günter
>
Kornel
signature.asc
Description: This is a digitally signed message part.
