PR 55905: ICE for polymorphic dummy argument with an allocatable coarray component

[Tobias Burnus]

Janus Weil wrote:
>> >and let me know
>> >the likelihood of a fix making it into stage 3?
> Zero (since stage three has already ended two days ago).
>
> To make it into the 4.8 release: Also close to zero (since it's not a
> regression: 4.7 gives me a couple of different errors on your test
> case).

I think there some leeway for simple fixes of bugs.

I believe the attached patch is correct - for a -fcoarray=single 
implementation.*

(For coarrays - in particular with -fcoarray=lib* - but presumably also 
for finalization, there are still some known issues with allocatable 
components.)

Tobias

PS: The attached patch is loosely tested. I intent to do a regtest and 
create a test-suite test for it.

* One issue with coarray components is that in an intrinsic assignment, 
they may not be reallocated, which currently happens. For 
-fcoarray=single, that's probably detectable using TARGET and a pointer. 
And for -fcoarray=lib, I am not even sure whether it properly calls the 
library (de)registering functions. That's an issue which is still high 
up on my coarray to-do list.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: class-coarray-scalar.diff
Type: text/x-patch
Size: 2463 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20130110/d9074aaa/attachment.bin>