[patch, fortran] Warn about suspicious assignment to contiguous pointers

[Janus Weil]

Hi Thomas,

>>> I think an unconditional warning is OK
>>> in this case because
>>>
>>> - Assigning to a pointer from an obvious non-contiguous target
>>>    is not useful at all, that I can see
>>
>>
>> I guess you're talking about a *contiguous* pointer here,
>
> Correct.
>
>> and in that
>> case I would argue that, beyond being "not useful", it's even illegal,
>> so why not throw a hard error, if we can infer at compile-time that
>> the target is non-contiguous?
>
> Problem is, we cannot infer this from the tests done.
> We would also have to add a test if the array is empty
> or that it contains only a single element, and that (I think)
> is a) impossible in the general case, and b) not worth the bother.

I'm not sure I understand which cases you're worried about here. Maybe
you can give an example?

In any case, I think your test case is a bit short, so I extended it
somewhat (see attachment) and found two cases along the way where your
patch throws a warning but shouldn't:

r => x(::-1)
r => x2(2:3,1)

The first one is a stride of minus one, which is valid and contiguous
AFAICS. The second one is obviously also contiguous.


>> As explained above, I'd vote for an error (or at least a conditional
>> warning, so that it can be disabled, like most(?) other gfortran
>> warnings).
>
> Attached is a version which makes this a warning enabled by -Wall;
> this should be enough to give people a heads-up.
>
> I have introduced most of your comments into the revised patch.

Thank you, looks much better already.

Apart from the two mishandled cases above, one other thing comes to my
mind: It might be a good idea to apply your checks not only to pointer
assignments, but also to dummy arguments (passing a non-contiguous
array to a contiguous dummy pointer), where the same rules should
apply.


> So, here is the updated patch.  Regression-tested on
> powerpc-linux, make dvi and make pdf also passed.
> OK for trunk?

Not quite, but we're getting closer :)

Sorry for being such a nag ...

Cheers,
Janus
-------------- next part --------------
A non-text attachment was scrubbed...
Name: contiguous_4.f90
Type: text/x-fortran
Size: 1341 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20170828/3805eaf5/attachment.bin>