On Fri, 4 Jan 2019 03:59:35 +0000 (UTC) "J. Smith" <etaion12-q...@yahoo.co.uk> wrote: > reassign 898550 dictionaries-common > stop > Hi, thanks for bringing this to dictionaries-common.
I am mostly out of my usual box, but will try to guess what is happening, sorry if it is too wild a guess. I guess we are dealing with hunspell-en-us Debian package. > Your backtrace indicates that the error comes from > debian-ispell-preprocess-dicts-alist, which is provided by > debian-ispell.el from the dictionaries-common package. Error is actually in `decode-coding-string' Debugger entered--Lisp error: (coding-system-error UTF-8) decode-coding-string("[']" UTF-8) ... > This is corroborated by the fact that the error disappears with -Q, which > prevents loading > of such site-lisp files. Perhaps the basic problem is using 'UTF-8 where > 'utf-8 is needed. That last seems to be the actual problem. Using -q does not load the initfile, but uses the registered value for hunspell-en-us which happened to have uppercase UTF-8 string. Using -Q will ignore everything and failure will not appear (no info at all about actual dict). For Debian hunspell dicts using uppercase string when registering (only hunspell-en-us AFAIK) this was noticed independently and fixed in dictionaries-common 1.27.4 shortly after this bug report was filed , https://salsa.debian.org/debian/dictionaries-common/commit/deece61f0068f20285d24d3a9ef9b3505cba748b But for the general use (personal config files and autodetected info from hunspell dicts) it is the config file or dictionary which should be fixed. There is nothing dictionaries-common can do here, although ispell.el could help by doing this lc conversion in ispell-get-coding-system. Not sure if there was a change in Emacs `decode-coding-string' around that time, that may explain why it worked before. Am I guessing too wildly? Regards, -- Agustin