[patch,gfortran] PR 21625: Initialization of derived types at allocation

[Paul Thomas]

Erik,

These are probably a naive questions:

(1) Why do you add the intialization in resolve_code rather than 
match.c(gfc_match_allocate)? The derived type is available at match 
time, isn't it?
(2) What is the business about making the last reference full?

You might just call the testcase derived_init_1.f90.

Apart from this, the patch looks fine, works OK and regtests on FC3/Athlon.

Paul T