[Bug-AUCTeX] Re: preview-1.248; preview completely broken?

[Richard Lewis]

David Kastrup <address@hidden> writes:

>> [The stuff below is with
>>
>> emacs -Q
>> `M-: (require 'tex-site) RET'
>> `M-: (load "preview-latex") RET'
>> `C-C C-P C-d'
>
> It helps if you are not reporting falsehoods.  The backtrace is for
> C-c C-p C-p, not for C-c C-p C-d.

yes indeed, sorry.

> Does C-c C-p C-d fail for you too?

sometimes...
before everything was failing, now it sometimes works.
One fool-proof way to reproduce seems to be:

mkdir m/
cd m
emacs -Q
M-: (require 'tex-site) RET
M-: (load "preview-latex") RET
C-x C-f blah.tex RET
C-c C-e document RET RET
<up> <up>
$a$
C-p C-p C-p
[y to save file and cache preamble]

If i open an existing file, then things sometimes work, and sometimes
dont, it seems to depend on the file and what is in the directory
(_region_ etc) and whether you try C-c C-p C-p first or C-c C-p C-d
first (and also the emacs session no doubt).

> This backtrace is pretty unsuitable for diagnosis.  Could you do
> M-x set-variable RET debug-on-signal RET t RET
>
> and then try again?  There might be some completely unrelated
> backtraces which you can skip by saying "c", but the backtrace with
> "preview-reraise-error" is too late for debugging.  The one
> immediately before it should be interesting.

with debug-on-signal the first backtrace produced is:

Debugger entered--Lisp error: (wrong-type-argument arrayp nil)
  coding-system-plist(nil)
  coding-system-get(nil alias-coding-systems)
  coding-system-base(nil)
  byte-code("ƒ