The following email had large attachments (more than 100K).  I uploaded
everything here:
http://mhbateni.com/context/bug-04/

On Thu, May 25, 2017 at 11:40 PM, Mohammad Hossein Bateni <bateni@gmail.com>
wrote:

> Hello Hans,
>
> I'm going to explain a strange problem which came about, as far as I can
> tell, in beta 2017.04.27 01:00.  The same beta had introduced some other
> issues with fallbacks, which you fixed already.  Thanks a lot for that!
>
> I am now using the latest beta, version 2017.05.25 12:50.  Unfortunately I
> don't know what is going on precisely.  Here is a sort-of-minimal
> non-working example.
>
>
> %%%%%%%%%%%%%%%%%%%%%%%%%
>
> \definefontfeature[slanted][slant=.4]
>
>
> \starttypescript [serif][sahel][name]
>
>   \resetfontfallback [dejavudefault]
>
>   \definefontfallback[dejavudefault][dejavusans]
> [range=generalpunctuation,check=yes,force=no]
>
> %
>
>   \definefontsynonym [SahelReg]    [Sahel][features={arabic,
> tlig},fallbacks=dejavudef]
>
> %% NOTE 1
>
> %  \definefontsynonym [SahelReg] [Sahel][features={arabic,tlig}]
>
>   \definefontsynonym [SahelBd]       [SahelBold][features={arabic,
> tlig},fallbacks=dejavudef]
>
>   \definefontsynonym [SahelIt]         [Sahel][features={arabic,tlig,
> slanted},fallbacks=dejavudef]
>
>   \definefontsynonym [SahelBdIt]     [SahelBold][features={arabic,
> tlig,slanted},fallbacks=dejavudef]
>
> %
>
>   \definefontsynonym [Serif] [SahelReg]
>
>   \definefontsynonym [SerifBold] [SahelBd]
>
>   \definefontsynonym [SerifItalic] [SahelIt]
>
>   \definefontsynonym [SerifBoldItalic] [SahelBdIt]
>
> \stoptypescript
>
>
> \starttypescript [math] [modern] [name]
>
> \definefontsynonym [MathRoman] [name:modern] [features=math\mathsizesuffix]
>
> \stoptypescript
>
>
> \starttypescript [myfont]
>
> \definetypeface [myfont][rm] [serif][sahel] [default]
>
> %% NOTE 2
>
> \definetypeface [myfont][mm] [math] [modern][default]
>
> \stoptypescript
>
>
> \setupbodyfont[myfont,rm]
>
> %% NOTE 3
>
> %\definedfont[Serif]
>
>
> \def\Sample{{\lefttoright Salam} {\righttoleft سلام} }
>
>
> \starttext
>
> rm:\Sample
>
> {\bf bf:\Sample}
>
> {\it it:\Sample}
>
> {\bi bi:\Sample}
>
>
> %% NOTE 4
>
> \definedfont[SahelReg]\Sample
>
> \definedfont[Serif]\Sample
>
>
> \stoptext
> %%%%%%%%%%%%%%%%%%%%%%%%%
>
>
> *Background*:
>
> Commenting/uncommenting some of the lines produces three outputs, only one
> of which seems correct to me.  I have attached the three.  Betas before
> 4/27 always produced the correct solution.
>
> I am using a free open-source Farsi font, whose English glyphs come from
> Open Sans font.  I am attaching the specific versions of the Farsi font to
> reproduce the same results.
> [https://github.com/rastikerdar/sahel-font]
> [https://github.com/google/fonts/tree/master/apache/opensans]
>
>
> *Explanation*:
>
> Compiling the MWE above produces wrong-1.pdf in recent betas.  The serif
> font has not been set up properly on the first line of output.  Instead the
> default Latin Modern has been used, which looks different and misses the
> Farsi glyphs.
>
> Strangely there is nothing wrong with the second line of text, which shows
> Serif and SahelReg fonts are both set up just fine.  See "NOTE 4" in the
> MWE.
>
> Replacing the line above "NOTE 1" with the line below it—i.e., if I remove
> the fallbacks for definition of SahelReg—generates correct.pdf as output.
> I get the same correct output if uncomment the line below "NOTE 3," which
> should be a no-op in my opinion.  Notice that setting the fallback for
> bold, italic and bold italic has not caused any issues.  It's only the
> serif itself that's acting up weirdly.
>
> Finally when I comment the math definition below "NOTE 2," I get
> wrong-2.pdf as output.  Here everything seems okay except that the serif on
> the first line—but strangely not on the second line—misses the Arabic
> features: the letters are not joined.
>
>
> I spent quite some time investigating the changes to the GitHub
> repository on 4/27, to pinpoint what might have caused this issue, but my
> efforts didn't go anywhere.  I hope you can spot the issue.
>
>
> Thanks a lot for your attention!
> —MHB
>
>
>