Description of problem: ispell/aspell fails from xemacs/emacs. After some debugging, cause is that ispell.el (see below *) calls for "-d portugues" to specificy the portuguese dictionary instead of "-d portuguese". In /usr/lib/aspell there's a portuguese.alias file but not a portugues.alias file (same for brazilian.alias, where a brasileiro.alias file may be needed.) Confront this with the files that come with aspell-fr where both french.alias and francais.alias are present. *actually the ispell.el (version 3.3.) shipped with xemacs doesn't even mention portuguese, but the latest stable available on the web (version 3.6) uses "-d portugues" as well as my .emacs file did before 3.6 was available. Also "ispell/aspell -d portugues" will fail from the command line, but I never use it that way. Version-Release number of selected component (if applicable): aspell-pt-0.50-6 How reproducible: always Steps to Reproduce: 1. run "aspell -d portugues -c <file>" 2. 3. Actual results: Error: The file "/usr/lib/aspell/portugues" can not be opened for reading. Expected results: file spelled Additional info: creating portugues.alias as a link to portuguese.alias (and also brasileiro.alias to brazilian.alias) in /usr/lib/aspell fixes it
Portuguese has a problem with its alias file in that it uses non-ascii in the filename-- and the non-ascii is not in UTF-8, which is required by RHL 8.0 and later. It (both rpm and aspell 0.50) did not like the conversion from iso-8859-1 filename to utf-8 (could not find it), so it was intentionally removed. The next version of aspell, which has UTF-8 support, will restore this functionality.
This problem has nothing to do with UTF-8. Emacs/Xemacs (in its ispell.el module) for some strange reason expects that Portuguese and Brazilian dictionaries will be named portugues and brasileiro. These are simple ASCII names. The aspell-pt dictionary really has a weird português.alias, but that's not the problem. The problem is the missing portugues and brasileiro alias (yes, brasileiro also, see bug #104843). Most modern programs that use aspell probably are using the ISO language code to especify dictionaries. Emacs/Xemacs don't. Emacs/Xemacs will be very happy if it worked ou of the box (even though I suppose they have been customizing it since the dawn of time :)). I'm creating an attachment with a replacement for the current aspell-pt-0.50-2.ascii.patch present in the source rpm. This patch removes the ISO-8859-1 alias and creates the ASCII ones. I have sent a similar patch upstream and I will see to it being applied. I would love to have this patch accepted for Fedora, or at least this bug reopened, as UTF-8 won't help here.
Created attachment 101305 [details] A patch to create the portugues and brasileiro alias.