[Patch, Fortran] PR41629: [OOP] gimplification error on valid code

[Janus Weil]

Paul,

> I'll look at this patch tomorrow.  I just got back from a trip.

thanks, that would be great. But if you do so, please don't look at
the patch I sent before, but at the update I'm attaching here.

In the previous version, I moved the call to
'encapsulate_class_symbol' to resolution stage only for plain CLASS
variables, but not for CLASS-valued components. I think it's better
(necessary?) to also do this for components. Note: For components the
attributes are known when parsing the component declaration (which is
just one line), but the type of the component does not have to be
previously defined. Also it's just cleaner to do the encapsulation at
the same point (namely: resolution stage) for *all* CLASS entities
(including components). So I moved 'encapsulate_class_symbol' to
resolve.c and made it static again.

As an aside, I also had to fix an error in gfc_match_allocate: There,
gfc_resolve_expr was called for the SOURCE tag, but of course
resolving should only happen at resolution stage, and not yet when
parsing. Consequently, some error checks (including the function
'conformable_arrays') had to be moved to resolve_allocate_expr, since
they rely on the SOURCE expression being resolved.

I think the patch is regression-free (will re-check). Of course I will
also update the ChangeLog.

Ok for trunk?

Cheers,
Janus
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pr41629_new.diff
Type: text/x-diff
Size: 22706 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20091010/f3b475c0/attachment.bin>