GCC's Unicode use [was Re: Compile performance of Linux kernelsin mainline gcc]

[Paul Schlie <schlie at comcast dot net>]

> Giovanni Bajo
>>Joseph S. Myers wrote:
>>> The fact that we now use UNICODE quotes should be specified in
>>> changes.html
>>> though, as a heads-up in case other users are misconfigured like you.
>>> Joseph, as the author of the original patch, would you please take care
>>> of this?
>>
>> Once the issue that the testsuite harness doesn't set LC_ALL=C like it
>> should has been dealt with (I think Mark was working on doing this
>> properly, in the .exp files rather than in the Makefile so it works for
>> people running runtest manually) so we know that issue doesn't need
>> documenting for 4.0, I'll add that documentation, referencing Markus
>> Kuhn's page <http://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html> as the
>> standard description of quoting issues for those not familiar with them.
>> (In particular,
>>
>> #   If you
>> #   can use only ASCII's typewriter characters, then use the apostrophe
>> #   character (0x27) as both the left and right quotation mark. If you can
>> #   use Unicode characters, nice directional quotation marks are available
>> #   in the form of characters U+2018 and U+2019.
>>
>> from that page explains the policy used: if you tell GCC that you can use
>> Unicode characters by your LC_CTYPE setting, you get the proper quotes,
>> otherwise you get neutral vertical quotes.)
>
> Very nice thanks! Please, make sure to also have a quick'n'dirty help for
> people having the same configuration problems of Andy's: changes.html after
> all is our little FAQ for people upgrading the compiler, and it often offers
> workarounds for problems in people code (see the long list of C++
> suggestions in 3.4). Thus, saying something along the lines of: "if you see
> garbage, check your LC_*, maybe it is set to UTF8 etc." will surely help
> diminuishing the number of invalid bug reports we will have to deal with.

Maybe GCC's generation of Unicode friendly quoted message text should be
explicitly requested as a command line option; thereby eliminating any
potential confusion over expectations by either the user, and/or non-Unicode
friendly programs which may subsequently parse GCC's output for whatever
purpose?

(However wouldn't expect GCC to ever accept or generate anything other than
ASCII quotes within text meant as code, as that would otherwise seem counter
productive.)