Re: [Patch, fortran] PR29396 PR29606 PR30625 and PR30871 - subreference array pointers.

[Tobias Burnus <burnus at net-b dot de>]

Paul,

Paul Thomas wrote: 
>> I saw none of that - I think that something bad has happened with
>> your trying to apply the secondary patch.  Note that this latter has
>> nothing to do with subref array pointers and is a regression due to
>> the patch that uses memcpy in assignments, where possible.
> I still think that this is the case.  I recharged my tree with the
> patch last night and regtested - all was well.
I removed the unrelated patch (and checked the result with interdiff),
but I still get the segmentation faults.

While I don't get a segfault for alloc_comp_constructor_1.f90 and
allocatable_dummy_1.f90 anymore, I continue to get one with PR30871.
Please recheck also with valgrind; valgrind shows here:

==19434==    at 0x4985B6: gfc_get_expr_charlen (trans-expr.c:210)
==19434==    by 0x487115: gfc_conv_expr_descriptor (trans-array.c:4614)
==19434==    by 0x49CF4D: gfc_trans_pointer_assignment (trans-expr.c:3465)

Tobias

PS: I heard that the patch submittal date and not the check-in date
counts - from this point of view, your patch is still stage2.
Additionally, it can be seen as bug fix which is allowed for stage3 (an
ICE is always a bug).