On 2016-03-09, Scott Kostyshak wrote:
> [-- Type: text/plain, Encoding: quoted-printable --]
> On Sat, Mar 05, 2016 at 09:36:43PM +0000, Guenter Milde wrote:
>> After all, this is a know upstream bug, not a "real" LyX issue (although LyX
>> may add a workaround as suggested in #8035).
> Do you have a link to the upstream bug report?
As said in #8035, this is no "official" bug, there is no report and I don't
know whether the polyglossia people would say it's a feature...
> So to make sure I understand, it would not be helpful if I sent an email
> to the TeX Live list with a minimal example of a document that succeeded
> before the update and now fails?
Not really. It it only fails with some fonts - actually only with some
versions of fonts, to it is hard to reproduce. Your minimal example may work
at some place and fail at another one. (Just like our tests did.)
Also, it is not a new problem but a long standing issue.
So, what is the problem:
1. LaTeX reports missing characters as a *warning*, not an error.
2. Polyglossia reports an *error*, if
* a (non-Latin) script is required by a language used in the document
* but the fontspec-configured font does not contain a "supports script
..." flag.
Test 1 is reliable and important (data loss):
* LyX converts the warning into an error. (Fixed)
Test 2 is unreliable:
* some fonts support scripts without the tested tag
* some fonts partially support a script
(e.g. a font may say: Greek is supported but lacks accented
characters in the "Greek Extended" block. On the other hand, Latin
Modern contains capital Greek letters but not small ones.)
* a language may be used without any text in the default script of
this language.
--> LyX should treat the error as a warning -> #8035
A possible bug report to polyglossia would suggest to issue a warning
instead of an error if a required script is not explicitely supported by the
font.
Günter
