Description of Problem: This bug is likely related to #64019 which got "CLOSED WORKSFORME". This time I have a much smaller sample, attached, which demonstrates the problem independently of settings of LANG. Results are the same regardless if emacs is run on a console or in a graphical enviroment and calling emacs with LANG="X" emacs sample.tex where "X" any of en_CA.UTF-8, en_US.UTF-8, en_CA, en_US and C does not seem to have any influence on outcome. If you will try 'ispell-buffer' on 'sample.tex' a process will bomb out with ispell-process-line: Ispell misalignment: word `lambdy' point 1823; probably incompatible versions There is no word `lambdy' at point 1823 although there is such label nearby. There is a slight difference if you will try the same thing using RH 7.3 distro ("UTF-8" was omitted in LANG settings). Namely a bomb message is even less "congruent with reality" and mysterious ispell-process-line: Ispell misalignment: word `lt' point 1598; probably incompatible versions The "real" ispell never had problems of that sort and was actually better in handling LaTeX markup. I tried the above on two different machines with 7.3 distro and on one with 8.0. Results were consistent. There are no troubles if "ispell sample.tex" is run and the file sports _one_ actual error (transposed letters in a word "strengthening"). But a workaround of that sort is a serious PITA if one works on a paper in emacs.
Created attachment 83566 [details] a file to demo a mismatch between aspell and emacs
I have a bit more. A LaTeX file which consists of such code %%%%%% for any $1 \le \ell \le k \le n$. Combining (\ref{www-ellips}) and (\ref{lambdy}) % and (\ref{lambdy-section}) %%%%%% on 8.0 system is causing this message after (ispell-buffer) in emacs: ispell-process-line: Ispell misalignment: word `lambdy' point 124; probably incompatible versions On 7.3 installation things are a bit more "creative". With this %%%%%% \begin{equation} \label{www-ellips} W({\cal E}, k, \ell) = \Bigl(\prod_{i =k - \ell +1}^k \la_i({\cal E}) \Bigr)^{-1/\ell}, \end{equation} %%%%%% the same (ispell-buffer) ends up with: ispell-process-line: Ispell misalignment: word `sgml' point 46; probably incompatible versions Where `sgml' is coming from I do not know. Most likely from aspell. I have also seen situations where aspell process goes away and never returns causing the whole command to block but I do not know how to reproduce that reliably. It seems to happen at random. In both cases removing '-' from 'www-ellips', or replacing it with '_', makes the command to complete without incidents. Replacing version 3.4 of ispell.el with a newer one 3.5 does not seem to change anything.
tested on Redhat 7.3, emacs-21.2 On about 70% of my computers aspell systematically hangs emacs in Latex mode on a file containing e.g. a line like that: \vspace{ This does not happen for root or if the user name is exactly one character long.
aspell will be upgraded in the next release