http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50640
--- Comment #19 from paul.richard.thomas at gmail dot com <paul.richard.thomas
at gmail dot com> 2011-11-07 11:00:47 UTC ---
Dear Tobias,
I have been keeping out of this for the time being because I want to
get some of the final OOP array issues out of the way. I figured
that, as a regression, it could wait for a little while :-(
>> Possible patch which pushes the vtab to the toplevel
This looks like a perfectly good way to do this.
>
> At least that approach fails for the following file (of PSBLAS 3's configure),
> which yields an
> internal compiler error: in cgraph_analyze_functions, at cgraphunit.c:1193
>
> which is at
> fprintf (cgraph_dump_file, "\nReclaiming functions:");
> for (node = cgraph_nodes; node != first_analyzed; node = next)
> {
> ...
> gcc_assert (node->analyzed == node->local.finalized);
>
> program xtt
> type foo
> integer :: i
> end type foo
> type, extends(foo) :: new_foo
> integer :: j
> end type new_foo
> class(foo), allocatable :: fooab
> type(new_foo) :: nfv
> integer :: info
> allocate(fooab, source=nfv, stat=info)
> end program xtt
That it should fail on this is utterly bizarre. I presume that it is
the allocate statement that wipes it out?
Paul
