Re: Patch to implement C99 flexible array member constraints

["Joseph S. Myers" <jsm28 at cam dot ac dot uk>]

On 17 Aug 2002, Gabriel Dos Reis wrote:

> | This is following the style of the surrounding code (some of which you
> | quoted). 
> 
> This is not a style issue.  
> Around the same code, you'll notice use of pedwarn(), so -not- using
> pedwarn_with_decl() in new code should not be an issue.

The universal current style in the C front end is that none of the
diagnostic features you mention (specifiers such as %D and %H) are used.  
The surrounding diagnostics are uses of error_with_decl.  Use of those
features (generally, not piecemeal) would be an improvement - but the C
front end is presently using the old style (and the C++ front end the new
style with just a few _with_decl uses).

> The {warning, error, pedwarn}_with_decl are intended to go out once
> mainline is again unfrozen for non-bug-fix patches -- I didn't
> unfortunately have the time to make all the removal before last 15th.

I see no sign that stage 3 has in fact yet started; though the tentative
date has passed, the web page still says stage 2 is in progress, and there
hasn't yet been an announcement to say that stage 3 has started.  And in
any case you seem to be saying that use of the _with_decl versions is a
bug.

> I repeated so last time I fixed misuse of it.
> I repeated it when I introduce the location format specifier; 
> I repeated when I was asked about the intent of introducing location_t.

Repeating is hardly good as documentation (nor are list messages;
documentation should be in CVS, editable and updatable).  (The proper
place would be a chapter in the internals manual about the diagnostics
interfaces, including the interfaces that are used by front ends to add
format specifiers.)

-- 
Joseph S. Myers
jsm28@cam.ac.uk