Re: Compile the parallel package using clang

[Sebastian]

On 2016-08-11, Olaf Till <address@hidden> wrote:
> 
> On Tue, Aug 09, 2016 at 10:20:08PM +0200, Olaf Till wrote:
>> On Tue, Aug 09, 2016 at 07:44:06PM +0200, Sebastian wrote:
>>> On 2016-08-09, Olaf Till <address@hidden> wrote:
>>>> 
>>>>> Testing for call-stack.h is probably a better idea: it fails as
>>>>> expected on 4.0.3 but succeeds on 4.1.0+.
>>>> 
>>>> Configure sees interpreter.h, but mkoctfile doesn't; and the issue
>>>> doesn't occur with call-stack.h? I see no obvious logic in
>>>> this. Better we try to track down the reason for the issue.
>>> 
>>> I am not an autoconf expert, however could there be another
>>> interpreter.h? (llvm has an interpreter.h) You check for
>>> existence, right?
>> 
>> Yes, you are right, that is probably the reason. (And the output you
>> attached looks ok so far.) octave-config returns an OCTINCLUDEDIR
>> which contains the final .../octave, and I had used this raw, calling
>> for this kind of trouble. I have now stripped the final /octave with
>> sed and tested for octave/interpreter.h (instead of interpreter.h),
>> which should solve the issue. But I've run into some mysterious bug in
>> my configure.ac code which I couldn't resolve up to now, so
>> unfortunately I can't send you a new configure.ac for testing as
>> yet. And I probably won't have time for this later today and in the
>> next few days ... sorry ... But I'll come back to it and fix it.
> 
> Done (hopefully). A new revision is pushed (there are several files
> affected). Thanks for helping to track this down.
> 
> Olaf

Yes, latest the revision works. I checked with 4.1.0+ and 4.0.3, both compiled 
by clang. 

Thanks,
Sebastian