Re: [Patch, fortran] PR24789 - ICE when assigning to array of strings

[Paul Thomas <paulthomas2 at wanadoo dot fr>]

Jakub,

> IMHO the type should be created already with the proper length known
> (whether it is the hidden _foo length PARM_DECL, or .foo artificial variable
> that is initialized in gfc_trans_deferred_vars for CHARACTER variables
> and dummy arguments and return vars), or for array types the various
> artificial vars initialized in gfc_trans_deferred_vars as well (except
> for assumed size arrays which of course don't have known size).
>  
I agree with that; automatic character length, dummy pointer arrays skip 
past all the places where they could be locally declared. This is 
remarked upon in one or two of the comments in trans_array.c and 
trans_decl.c as if it is completely evident that this should be so. They 
depend entirely on the actual argument backend declaration, which is 
missing the TYPE_UNIT_SIZE exactly because the character length is 
automatic.

I guess that the right thing would be to create these variables locally. 
For the main part, gfc_trans_deferred_vars is doing the job of 
initialising the character length; that was Erik's contribution to the 
patch. It is just that the lack of a local declaration means that the 
element type size information is missing, so that those parts of 
gfortran that still use this segfault.

> I can distill non-OpenMP dependent changes from the gomp branch soon
> for review if there would be interest in that.
>  
Looking through your patches, I realise that this is something very 
important to incorporate into trunk. The incomplete types need to be fixed.

Sorry about the slightly intemperate response last night. I took a silly 
amount of spare time to fathom out what was going on and to come up with 
a less than perfect patch.

Best regards

Paul