Re: [Patch, fortran] PR78737 - [OOP] linking error with deferred, undefined user-defined derived-type I/O

[Janus Weil <janus at gcc dot gnu dot org>]

2016-12-13 10:42 GMT+01:00 Janus Weil <janus@gcc.gnu.org>:
>> please find attached a version of your patch that runs all the dtio
>> testcases successfully.
>
> Great, thanks a lot. Your addition of
> gfc_find_and_cut_at_last_class_ref is just what I was looking for
> right now ...
>
> If you don't mind I'll write a ChangeLog, add the proper tests and
> commit to trunk. Or do you prefer to take care of it yourself?

Btw, to continue the brainstorming, I think I found a slightly better
solution for the dtio_13 problem, which even removes the spurious
error on that test case via a small fix in resolve_transfer. The
attached patch is what I'd like to commit if you're ok with it.

Cheers,
Janus



>> On 13 December 2016 at 06:33, Paul Richard Thomas
>> <paul.richard.thomas@gmail.com> wrote:
>>> Dear Janus,
>>>
>>> I woke up in the middle of the night realising that not only are you
>>> right about the need for dynamic dispatch but that my dtio_20.f90 must
>>> already work.
>>>
>>> Thanks for putting me right!
>>>
>>> Paul
>>>
>>> On 13 December 2016 at 00:30, Janus Weil <janus@gcc.gnu.org> wrote:
>>>> Hi Paul, hi all,
>>>>
>>>> 2016-12-12 21:04 GMT+01:00 Janus Weil <janus@gcc.gnu.org>:
>>>>> As commented several times in bugzilla, my feeling is that the
>>>>> solution for this PR would be to utilize the vtable machinery, in
>>>>> order to generate a truly polymorphic call to the DTIO procedure.
>>>>
>>>> in order to elaborate what I have in mind, I'm attaching a draft patch
>>>> which implements polymorphic DTIO in the most straightforward manner I
>>>> could come up with. I have not regtested it yet, but at least it
>>>> removes the link failure on comment 0 and 6 in the PR and most
>>>> importantly it generates the correct output for comment 18, which none
>>>> of the previous attempts have accomplished.
>>>>
>>>> I'd be grateful for any comments, in particular whether I'm on the
>>>> right track here or whether I'm misinterpreting the F03 standard in
>>>> any way ...
>>>>
>>>> (Btw, it seems that Paul's dtio_20.f90 works already on current trunk,
>>>> so it's not very well suited to test for the problem at hand.)
>>>>
>>>> Cheers,
>>>> Janus
>>>>
>>>> PS: A quick check of the dtio_* tests shows ICEs on dtio_7.f90 and
>>>> dtio_13.f90. I'll look into those tomorrow.
>>>
>>>
>>>
>>> --
>>> If you're walking down the right path and you're willing to keep
>>> walking, eventually you'll make progress.
>>>
>>> Barack Obama
>>
>>
>>
>> --
>> If you're walking down the right path and you're willing to keep
>> walking, eventually you'll make progress.
>>
>> Barack Obama

Attachment: pr78737_v2.diff
Description: Text document