Re: [PATCH, Fortran] PR 33141: Better diagnostics and handling of intrinsics with -std=* settings

[Daniel Kraft <d at domob dot eu>]

FX wrote:
> Hi Daniel, hi all,

Hi FX,

I'll try to answer the questions despite the check-in and please send 
your comments for a small update patch if we agree:

> 1. What do -fall-intrinsics and -Wintrinsic-shadow do when used 
> together? I suspect their behaviour is close to that of 
> -Wintrinsics-std, isn't it?

No.  This combination will warn about a user-procedure having the same 
name as any intrinsic available, like:

FUNCTION asinh (arg)
         1
'asinh' at (1) is also the name of an intrinsic; explicit interface or 
EXTERNAL might be needed to call it (message sketched).

-Wintrinsics-std will warn on the *call* of an intrinsic that is not in 
the standard (thus, -fall-intrinsics will obsolete -Wintrinsics-std):

PRINT *, asinh(1.)
         1
Intrnisic 'asinh' is not in the standard and 'asinh' is treated as if 
declared EXTERNAL.

(Thus, for those two code-snippets combined, the FUNCTION will be called.)

> 2. Why is -Wintrinsic-shadow the only "intrinsics" option that has the 
> singular form used? I think the plural would be fine there.

Actually, I dislike -Wintrinsics-std more (as below); however, I believe 
-Wintrinsics-shadow-allstd is not what it does, see above.  And I 
couldn't think of a good name here, but am open for suggested changes :)

The singular here is, well, just my first idea of the name.  But I think 
it fits quite good, as a single intrinsic is shadowed...  But here I'm 
also open for name suggestions, of course.

> 3. I don't quite like the name of "-Wintrinsics-std", because I don't 
> see how it's linked to what it does. The one name I could think of is 
> -Wintrinsics-shadow-allstd, but other people might have ideas.

See above.

> Otherwise, your revised patch seems OK. Please remember to update the 
> "New features" page, as it's otherwise harder when the release comes to 
> write release notes without forgetting anything.

Done, at the moment with the option names as checked in.

Thanks for the comments,
Daniel