[cfe-commits] Implement 'mutable'

[Daniel Dunbar]

Such foresight! Using substitution doesn't preclude localization in
the source, but if you wish to avoid for this reason then ok. I
believe there to be quite a few other places in clang where we do do
some substitution in the source code of a hard-coded string.

At some point I think we need to have a discussion about what the
right way to handle diagnostics is. For example, how should warnings
be uniquely identified, what rules should their format / grammar
follow, what are the guidelines for which locations should be marked,
etc. This is probably a ways down the road though.

 - Daniel

On Thu, Nov 13, 2008 at 1:02 PM, Sebastian Redl
<sebastian.redl at getdesigned.at> wrote:
> Daniel Dunbar wrote:
>>
>> Not sure if this is what Doug meant, but instead of "splitting" this
>> is probably a good candidate to just replace the mouthful with %0 and
>> supply the appropriate value at run time.
>>
>
> Nah, 'reference' is not a language keyword, so the message wouldn't be
> localizable. I could make one diagnostic for non-const and non-static, but
> the static/mutable mismatch is detected by the parser, so the "non-static"
> violation is never reported by the sema.
>
> Sebastian
>