Re: [Peter Dyballa] Re: Coding system of file not recognised correctly

[Chong Yidong]

Peter Dyballa <address@hidden> writes:

> Am 07.10.2006 um 22:14 schrieb Jason Rumney:
>
>> Peter Dyballa wrote:
>>> The reason it *had*, is that I had in load-path four ELisp files:
>>
>> So if I understand you correctly, this was a local configuration
>> problem, and we can remove this bug from FOR-RELEASE?
>
> No, I wouldn't say so. In the extra directory were the non-compiled
> ELisp files. These files are stable, unchanged, since last year,
> obviously (I forgot to diff before I deleted them, seeing that the
> byte counts were the same). So GNU Emacs was and is loading the same
> contents.
>
> And by copying the un-compiled ELisp files into this directory I
> again get the first opened UTF-8 encoded file with -E in mode-line,
> only the second file has the correct -u. When I copy the byte- 
> compiled files, from last 'make bootstrap', also into /Library/
> Application Support/Emacs both files open at once with -u in their
> mode-lines ...

As far as I can tell, it basically boils down to "don't copy the CJK
translation table files to other places in your load path".  AFAICT,
these files are deliberately loaded at run-time by
`utf-translate-cjk-load-tables', and messing with them can mess up how
utf-8-post-read-conversion functions.

I suggest leaving this as it is, and possibly changing the system to
something more robust after the release.  In the meantime, just set
`utf-translate-cjk-mode' to nil if you want to mess with the
translation table files.