Re: [PATCH] support a few of the new features of C++11 in syntax highlighting

[Daniel Colascione <dancol@dancol.org>]

[-- Attachment #1: Type: text/plain, Size: 2926 bytes --]

On 11/18/2014 10:01 AM, Óscar Fuentes wrote:
> Daniel Colascione <dancol@dancol.org> writes:
> 
>> On 11/18/2014 08:36 AM, Rüdiger Sonderfeld wrote:
>>> On Tuesday 18 November 2014 15:47:43 Óscar Fuentes wrote:
>>>> Daniel Colascione <dancol@dancol.org> writes:
>>>>>>> The "alignof" keyword is the only one still missing.
>>>>>>
>>>>>> From the top of my head: override is also missing.
>>>>>
>>>>> Because "override" (and "final") are keywords only in certain semantic
>>>>> contexts, supporting them requires special care.
>>>>
>>>> override and final have the same context as `const' for methods, so we
>>>> could take that as a model. Skimming over cc-langs.el didn't show an
>>>> obvious place for it, though.
>>>
>>> Yes, but "const" has a meaning in other places as well.  That's why it is 
>>> dealt with in `c-type-modifier-kwds'.  I think the problem is similar for 
>>> "noexcept".  I've simply added it to `c-type-modifier-kwds' for now.  But it's 
>>> actually the wrong place.  However unlike "override" and "final", there is 
>>> also an operator version of "noexcept".
>>
>> C++11 introduced the noexcept and constexpr keywords. They're always
>> keywords. "override" and "final" are different. Please don't go hack
>> those into cc-engine. Better to leave them unfontified than to fontify
>> perfectly good identifiers as keywords.
> 
> Is this the problem? *If* hacking c-mode font-lock system ends with
> "override" and "final" fontified as specifiers, is that so terrible? Are
> those identifiers so popular? C-mode already gets confused very often
> about how to fontify certain non-C-ish chunks of C++ code, so having
> "override" and "final" fontified as specifiers everywhere seems a
> non-issue to me. What's more annoying is to not fontify them on method
> declarations, because my first reaction when glancing at that code is to
> assume a typo.

Patch cc-engine to recognize the proper context if you care so much. Until you can do that, do not touch cc-mode.
 
>> They're not like "const" at all.
> 
> As far as c-mode is concerned, they are like "certain variety of const."

No, lexically, they are completely different. You cannot have a variable called "const". You can have a variable called "final".

> If c-mode uses the same heuristics everywhere for fontifying `const',
> that means that we cannot exploit the existing mechanism for fontifying
> "override" and "final". To bad. Then I would vote for adding them to
> some list of keywords. As I said, not having them fontified when they
> should is worse than having them fontified when they shouldn't.

Do not introduce bugs into Emacs cc-mode. It's one of Emacs most widely used features and will not change to suit your non-universal preferences. If you want to err on the side of over-highlighting, you are free to create a derived mode locally.

Do it right or not at all. 


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]