# [tex-live] TL2018: UTF-8 characters in file names with pdflatex

Zdenek Wagner zdenek.wagner at gmail.com
Sat Jul 28 17:15:31 CEST 2018

Hi all,

in the meantime I tried a minimal example and it works so there must be an
incompatibility in one of the packages. I will examine it further when I
have time.

Zdeněk Wagner
http://ttsm.icpf.cas.cz/team/wagner.shtml
http://icebearsoft.euweb.cz

2018-07-25 12:39 GMT+02:00 Zdenek Wagner <zdenek.wagner at gmail.com>:

> Anyway, as I wrote, I had no time to test it and I will have no time at
> least to the end of September.
>
> Zdeněk Wagner
> http://ttsm.icpf.cas.cz/team/wagner.shtml
> http://icebearsoft.euweb.cz
>
> 2018-07-25 12:26 GMT+02:00 David Carlisle <d.p.carlisle at gmail.com>:
>
>> hmm for all variants that I can detect are using things incompatible
>> with inputenc it is not preloaded
>>
>>
>>
>> % Skip this section in Unicode TeX, or if  MLTeX and EncTeX are enabled.
>> %    \begin{macrocode}
>> \ifnum0%
>>   \ifx\Umathchar\@undefined\else 1\fi
>>   \ifx\mubyte\@undefined\else 1\fi
>>   \ifx\charsubdef\@undefined\else 1\fi
>>   =\z@
>>
>>
>> so in a format built with enctex there should be no change in the 2018
>> release (or that's a bug, sorry)
>>
>> David
>>
>>
>> On 25 July 2018 at 10:51, Zdenek Wagner <zdenek.wagner at gmail.com> wrote:
>> > Hi all,
>> >
>> > I was too busy with my own work so that I almost have not tested TL
>> 2018. I
>> > found that my LaTeX documents do not work and since I have all releases
>> > starting at 2014, I have just switched back and had no time to examine
>> it
>> > further.
>> >
>> > Now I probably know the reason. I often use constructions as
>> > \futurelet\next\dosomething or \everypar{\setinitial} with
>> > \def\setinitial#1{... do something with the first character ...}. For
>> these
>> > macros everything must be a character with category 11 or 12, not an
>> active
>> > character. The inputenc package in case of UTF8 will give me just a
>> part of
>> > the character. For this reason I use encTeX where active characters are
>> not
>> > needed. As I am thinking about it just now, it seems to be easily
>> solvable.
>> > UseRawInputEncoding\endcsname which will revert the encoding for new
>> LaTeX
>> > and insert harmless \relax in all other cases. The same encTeX files are
>> > also used in the plain TeX so \csname seems to me as most portable.
>> >
>> >
>> > Zdeněk Wagner
>> > http://ttsm.icpf.cas.cz/team/wagner.shtml
>> > http://icebearsoft.euweb.cz
>> >
>> > 2018-07-25 11:36 GMT+02:00 David Carlisle <d.p.carlisle at gmail.com>:
>> >>
>> >> On 25 July 2018 at 09:18, Zdenek Wagner <zdenek.wagner at gmail.com>
>> wrote:
>> >>
>> >> >
>> >> > pdflatex is an 8-bit SW, it is not Unicode-aware. It is achieved by
>> >> > active
>> >> > characters which break a lot of things. Almost all my older files
>> ceased
>> >> > to
>> >> > work with pdflatex and TL2018 .
>> >> >>
>> >>
>> >> That's very surprising, you could have reported at least one to the
>> team!
>> >>
>> >> Since the release I have only seen one report of a user file that
>> >> generated an error
>> >> (and that was silently mos-interpreting non ascii input in older
>> >> releases so the error wasn#t really a bad thing)
>> >>
>> >> If you have an example of a conforming document that failed to work
>> >> after the 2018 release let us know
>> >> at the very least we can add to the documentation to say what to do,
>> >> if that is not covered by the existing documentation.
>> >>
>> >>
>> >> David
>> >
>> >
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://tug.org/pipermail/tex-live/attachments/20180728/80c7b3f0/attachment.html>