Set LC_ALL=C for all usages of comm(1), sort(1), tr(1)

[sideeffect42]

For tr(1) I only set $LC_ALL when character classes or ranges are used.

All three utilities are dependant on the locale setting. And especially when locale can differ between the local system and the target, sorting order can vary, leading to unexpected bugs.
Also regex character classes and ranges can vary depending on locale.

[sideeffect42]

Hmm, now that I'm thinking about it, wouldn't the locale selection also affect globs?

[4nd3r]

wouldn't the locale selection also affect globs

Sort order.

Want do that separately? Approved the PR, tho.

[sideeffect42]

Want do that separately?

Question is what to do in the case that we would have to set the locale for all globs.
I think this would be going overboard and I was thinking about maybe setting the locale to C in skonfig instead, for all scripts.
I think using the POSIX locale is what most people expect and if some other locale is required it can always be overridden.

I think I need to set up an Estonian VM to test this a bit 😃.

[sideeffect42]

Quick test suggests that only the Korn shell (not MirBSD Korn Shell (mksh), though) respects the locale in globs.
Though for Estonian A-Z also matches T U V W X Y Z. 😕.
bash, dash, zsh seem to treat globs as in POSIX locale always.

Test e.g.:

# alternatively: LC_ALL=et_EE.UTF-8
LC_ALL=de_CH.UTF-8 ksh -c 'for c in A B C D E F R S T U V W X Y Z Ä Ö Ü; do case $c in ([A-Z]) echo "$c";; esac; done'

[4nd3r]

setting the locale to C in skonfig instead

I think we should do this.

[sideeffect42]

setting the locale to C in skonfig instead

I think we should do this.

Yes, the more I'm thinking about this problem I also come to this conclusion.

I close this PR and then try to modify skonfig accordingly.