wacom-hid-descriptors
wacom-hid-descriptors copied to clipboard
linuxwacom
force LANG=C on fprint to ensure dot as decimal-sparator
ran into this when using the script on a german localed fedora which is using , as decimal separator.
In that case, potential error messages would also be in C instead of the one from the user. another problem is the encoding. If it is applied on the whole script, C.utf8 should be used—but works only in utf8-environments then.
The script itself is not multilingual: status updates, errors, and documentation are all in English. I'm not that concerned with keeping subprocesses multilingual. The sysinfo script uses LC_ALL=C and we haven't any any language or encoding complaints to speak of.
Comparing LC_ALL and LANG, the former seems like the better option since the latter is still subject to user-specific overrides that we're trying to avoid. Also, the docs indicate that we might be able to set both LC_ALL=C and LANGUAGE=$LANG to get a mix of both worlds: C-locale semantics with non-English messages. That combination isn't going to be perfect, but maybe closer to your ideal?
Oh, I don't have an ideal :smile:. I only try to point out the pros and cons :wink:. The script could also set LC_NUMERIC for this specific problem. You won't get a perfect solution in any way. A „perfect“ solution would be that everything works in the locale (i.e. $PEN_DATA contains localized values) or normalized (i.e. fprint accepts dot even on de_DE).
My two cents: Playing with mixed locale-environments is always the same f***up like when working with unnormalized, different timezones. LANG is indeed not a good solution, the script should either LC_NUMERIC=C for the single fprint-commands or set a LC_ALL in the header. for LC_ALL I'd use en_US.utf8 instead of C.
What do you want? :smiley:
I agree with @Blaimi, mixing two locales is bound to cause more issues. Let's force LC_ALL=C and let's see if anyone complains.
As for C.utf8 I'm not sure I have enough opinion here, does it matter for what we're doing here?
Works for me. Please just set export LC_ALL=C at the top of the script for now. We can explore if UTF-8 is necessary in another request since the same logic will probably also apply to sysinfo.sh.