On 2016-01-07, Scott Kostyshak wrote:
> On Thu, Jan 07, 2016 at 06:31:11PM +0100, Kornel Benko wrote:
>> Am Donnerstag, 7. Januar 2016 um 17:22:09, schrieb Guenter Milde
>> <mi...@users.sf.net>
>> > Maybe we should add a pattern
>> >
>> > export/(doc|examples)/he/.*pdf5.*
>> >
>> > to "unreliableTests" for now.
>> Wait a moment. In an another (luatex) list Scott is investigating too.
>> Looks like some defines would be sufficient to compile with lualatex
>> too.
>> % prior to specifying document class
>> \let\luatexpardir\pardir
>> \let\luatextextdir\textdir
>> \documentclass[english,hebrew]{article}
> Indeed. The relevant thread is here:
> http://tug.org/pipermail/luatex/2016-January/005591.html
> I don't have an opinion as for whether to change the .lyx file or to
> treat the test as unreliable.
I don't think this can be fixed in the .lyx file, because I don't know of a
way to specify something to be printed before the \documentclass line!
Generally,
* if the example/template/documentation source can be made more robust
without becoming "hackish" --> fix the source
* if the test machinery would need to be changed to "massage" a source to
compile --> invert the test case
* if LyX could be enhanced to care for a permanent TeX limitation
--> file a ticket at track and invert the test case (inverted:lyxbug)
* if a temporary TeX issue could be fixed by special-case handling in
LyX but is rather a corner case --> invert the test case (inverted:texissues)
Specifically:
* I don't think this merits special-case code in LyX to prepend these
lines if Hebrew is compiled with luatex.
The above linked message says "*until the package is updated* you can
simply alias the old name, ...". So we would add a hack for a temporary
failure with a non-supported output format (polyglossia says
Hebrew+LuaTeX is unsupported).
* FreeSans seems to provide the correct script-support info not to trigger
the false-positive error http://www.lyx.org/trac/ticket/8035
However, setting all of "mainfont", "sansfont", and "monofont" to FreeSans
seems "hackish" to me.
> So whatever you guys decide is fine with me.
My suggestion depends on how you (Scott) get the Hebrew documents to compile:
* special code in the test machinery (replacements or additions to the LyX
source or the exported .tex file),
-> remove special code and re-activate the patterns in "suspiciousTests"
* you have an old TeXLive version where this works
-> move patterns to "unreliabelTests"
* it does not longer work after updating TeXLive
-> re-activate the patterns in "suspiciousTests"
In any case, I propose to replace the special code setting system fonts
in the test scripts with "non-hackish"¹ system font settings in the
sources (for manuals this requires Uwe's approval).
¹ For Hebrew, this means that if there is no working free serif font, we
should invert the affected tests as "lyxbug" #8035 instead of using
a sans-serif font as serif!
Alternatively, we could try if the examples work with Babel (and
produce correct output) and set "always Babel" in the sources.
Thanks,
Günter
