From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.7.3) Gecko/20040913 Description of problem: When editing any .tex file, hitting M-$ (or any ispell function) hangs the process (C-g can recover from it). M-x flyspell-mode hangs the process in a way not recoverable from C-g. The symptoms disappear if one disables the Ispell TeX parser: (add-hook 'tex-mode-hook (function (lambda () (setq ispell-parser nil)))) (add-hook 'TeX-mode-hook (function (lambda () (setq ispell-parser nil)))) (add-hook 'LaTeX-mode-hook (function (lambda () (setq ispell-parser nil)))) Version-Release number of selected component (if applicable): emacs-21.3 How reproducible: Always Steps to Reproduce: 1.Open a .tex file 2.Hit M-$ on any word in the text (say, APPLICATION) 3. Actual Results: Emacs shows "checking APPLICATION..." and hangs. Expected Results: It should tell me the word is correct and go on editing. Additional info:
This sounds a bit like bug 54391, which I couldn't reproduce... Do the same thing happen with "emacs -q --no-site-file" for you?
Perhaps it is a aspell issue on x86_64? Can you reproduce on a ix86 box? :) Perhaps you'd like to try with say FC3test3 or the packages in rawhide now to see if it is any better.
The same problem happens with "emacs -q --no-site-file" on the AMD64 machine. I do not have a non-AMD64 x86 box with Fedora Core 2 installed. This is a workplace machine, it's not easy for me to go and install experimental machines.
I can confirm the hang on AMD64 and working on 32-bit Athlon. Steps: $ touch empty.tex $ emacs empty.tex type "barf" hit ESC hit $ Result on single Athlon XP 2400+: it nicely confirms that the word "barf" is correctly spelled. Result on dual Opteron 248: prints 'Checking spelling of BARF...', then emacs hangs with 100% CPU usage. Both machines are running Fedora Core 2, package versions emacs-21.3-12, aspell -0.50.3-19.1. Opteron is running 64-bit version. If I edit 'empty.txt' instead of 'empty.tex' the Opteron does not hang.
emacs loops on a read(3, "", 1024)=0 system call. File descriptor 3 is a pipe (read end) to ispell, set to non-blocking mode.
Ok, I see. Does FC2 aspell work ok by itself on amd64?
It looks like aspell works well standalone. At least, I can launch it on the same file and it works ok.
The problem does *not* appear on Debian pure64 with the following packages: emacs21 21.3+1-5.0.0.1.pure64 The GNU Emacs editor aspell 0.50.5-4 GNU Aspell spell-checker
This seems to be an aspell issue as I suspected: Downgrading to aspell-0.50.3-19.1 (fc2) on an fc3 install and I can reproduce the problem.
(I forgot to add that it works in FC3.)
Can anybody reproduce this bug with the latest aspell package (aspell-0.50.5)? Ivana Varekova
I'm closing this bug. If there is some problem please reopen it.