[Patch, fortran] PR19292 - g77 features missing or broken: PR18737 and PR21565

Paul Thomas paulthomas2@wanadoo.fr
Wed Oct 26 14:34:00 GMT 2005


Tobi,

>There are a lot of places where we call gfc_set_default_type with error_flag =
>0, are you sure that issuing the errors is correct in all these places?  If
>that's not certain, a safer approach would be changing the callsite in
>resolve.c to only set error_flag == 0 if the symbol is an intrinsic function.
>  
>
One thing that I am sure of is that a failure, anywhere, to find a 
default type for an external symbol is always an error.  I am not so 
sure that doing it your way round is always correct.  However, I will 
retire for a few quiet moments this evening to contemplate the issue.  
It's time I updated the Doxygen documentation isn't it?

>>2] PR21565: Add to check_conflict the constraint that a namelist object
>>cannot be in block data.
>>    
>>
>
>This is of course ok.
>  
>
Thanks

Paul



More information about the Gcc-patches mailing list