Hi GFortraneers,
I like to present a small series of patches. While working of PR90076 and
figuring how to best set the _vptr of class types, I discovered several ways of
doing this in slightly different ways which are more or less complete (mostly
rather less). I therefore decided to fix not only the PR90076, but also to
approach the issue in a more general way. This results in these three patches:
1/3 Fix PR90076 by applying gfc_set_vptr() where missing.
2/3 Rigorously apply gfc_set_vptr() (and extend it slightly) to reset the _vptr
consistently.
3/3 RFC: First step on assigning to the _vptr.
Working on the second part of the patch series, I figured, that at several
locations in the fortran compiler we are assigning to class data types and each
location does it differently (more or less complete). Therefore I got the idea
of creating a general function to do all necessary steps for assigning to a
class data type, like for example `gfc_add_modify_class (gfc_se *, tree to,
<something> from)`. The gfc_se is chosen to be able to produce a `pre` and
`post` set of statements, while the _data assignment is done in the `expr` to
be able to use gfc_add_modify_class() by the scalarizer.
I understand that this is long way to go. Therefore I first like to ask about
the general perception of this idea. So I like to hear constructive criticism.
Could I explain it good enough? Do you think it is worth pursuing?
Regards,
Andre
--
Andre Vehreschild * Email: vehre ad gmx dot de
