[Patch, Fortran] PR41583 - TYPE IS rejected because of way vtable index is implemented

[Dominique Dhumieres]

> The drawback, of course, is the unlikely possibility of collisions
> between the hashes for different derived types within a given class.

My (limited) understanding of hash tables is that they are supposed
to handle collision, i.e., if a collision is detected, the hash function
is repeated until an empty spot is found (write) or there is a match
(read).

> Perhaps this could be solved by an error, saying that one of the names
> should be changed?

If such a detection is done (IMO it should) then letting the user find
a name not colliding with others will be very frustrating for very big
codes. I think the only viable solution is to implement the process
outlined above (repeat the hash function).

BTW is 100000000 the number of entries in the hash table?

Cheers,

Dominique