Re: commit-msg hook

[Paul Eggert <eggert@cs.ucla.edu>]

On 04/14/2015 10:09 AM, Eli Zaretskii wrote:
>> This sort of thing should work in a unibyte environment, but it needs to
>> >be used only after testing that we actually are in a unibyte environment.
> I thought that's what all the tests with cent_sign and at_sign do,
> don't they?

No, they test for something more specific, namely, whether we're in a 
UTF-8 locale.  Not every multibyte locale uses UTF-8.

> what bad things could happen if this regular expression is
> used in a multibyte environment?
I suppose it could cause the script to print "Unprintable character in 
commit message" even though all the message's characters are actually 
printable.

How about this idea?  Before falling back to the unibyte regular 
expressions in awk, set LC_ALL='C' in the environment.  This should work 
well enough, as in practice all environments where the C locale is 
multibyte have working UTF-8 so they won't need to fall back to unibyte 
anyway.