Re: [Patch, Fortran] PRs 43062/47339 NAMELIST fixes

[Tobias Burnus <burnus at net-b dot de>]

On 01/26/2011 10:53 AM, Paul Richard Thomas wrote:
> I wonder how we have managed to get up to NAMELIST_70.f** ??? :-)

I  think mostly due to libgfortran-related stuff. Especially with arrays 
and derived types you have a huge number of possibilities how a user can 
enter a namelist, which is not really helped by the fact that gfortran 
also has some vendor extensions. Namelists with defined I/O for 
polymorphic or derived-type entities will give the next opportunity to 
introduce bugs.

> Given that the correction to the standard still needs to get passed in
> a WG5 ballot, should we not produce a warning with -pedantic??

Well, the standard is ambiguous thus some edit is needed. However, 
looking at the standard, the intent is rather clear that 
allocatables/pointers should be allowed. And the current wording only 
prohibits at most local pointer/allocatables - not already 
allocated/associated host-/use-associated or dummy 
pointers/allocatables. Thus, I doubt that WG5 will change the 
interpretation.

> OK for trunk.  Many thanks for wading through the interpretation
> request process and getting a clear response.

One should also thank Steve for spotting the ambiguity in the standard.

Thanks for the review. Committed as Rev. 169282.

Tobias