Re: [Patch, fortran] PR102619 - [13/14/15/16 Regression] ICE in gfc_conv_descriptor_dtype, at fortran/trans-array.c:215 since r9-6493-g0e3088806577e805

Hi Paul,

this looks very localized and concise so that I think it is
good not only for mainline, but also for a cautious backport.
(Let it sink in for a week before doing this).

Thanks for the patch!

Harald

On 3/27/26 14:09, Paul Richard Thomas wrote:
> Hello All,
> 
> As pointed out by Mikael in the PR, this bug was due to interface
> mapping not using a descriptor type for an assumed rank formal of a
> non-intrinsic function. The attached patch remedies this.
> 
> Bootstrap with FC43/x86_64. OK for mainline and backporting?
> 
> Cheers
> 
> Paul