[tex-live] Fwd: Duplicate Thai patterns reported by XeTeX in TL 2015 pretest

Mojca Miklavec mojca.miklavec.lists at gmail.com
Wed Apr 15 15:10:25 CEST 2015


On Wed, Apr 15, 2015 at 10:17 AM, Jonathan Kew wrote:
> On 15/4/15 02:01, Dohyun Kim wrote:
>
>> This issue might be related to the weird behaviour of xetex I have found:
>>
>> \toks0{유
>>>> }\showthe\toks0
>> \bye
>>
>> $ xetex test.tex
>> This is XeTeX, Version 3.14159265-2.6-0.99992 (TeX Live 2015/dev)
>> (preloaded format=xetex)
>>   restricted \write18 enabled.
>> entering extended mode
>> (./test.tex
>>>
>>>   \par .
>>
>> l.5 }\showthe\toks0
>>
>> ?
>>   )
>> No pages of output.
>> Transcript written on test.log.
>>
>> As shown, characters `유' and `술' are gone away and nothing is printed.
>> Note that Unicode codepoints of these characters are U+C720 and U+C220
>> respectively, with 0x20 in their lower bytes.
>>
>
> Note also that in the original report about Thai patterns, the error
> messages show lines from the patterns file with their last character U+0E20
> (ภ) missing. Again, 0x20 in the lower byte (which will be the leading byte
> in the buffer on little-endian platforms).
>
> It looks to me like there could be something broken fairly early in the
> input-scanning process (but after conversion from UTF-8 to UTF-16) whereby a
> line-final UTF-16 character that begins with 0x20 is being discarded as
> though it were a <space>. But this apparently doesn't affect the Win32
> binary, as the testcases work fine for Akira. (Is it ONLY on OS X, or has
> this been observed on other platforms?)

After some very very basic testing:

- There is a problem with the latest XeTeX as shipped with TeX Live
2015 pretest on Mac OS X.

- I tested x86_64-linux, sparc-solaris and Kakuto-san tested windows.
There are no problems on those platforms (using exactly the same TeX
support files). The x86_64-darwin binary from TeX Live 2014 also works
as expected, so it's unlikely that there would be something
fundamentally wrong with (La)TeX support files.

- The problem is reproducible with both x86_64-darwin and i386-darwin
(universal-darwin).

Mojca



More information about the tex-live mailing list