[m-rev.] for second review: Add future datatype for concurrent and parallel programming

[Julien Fischer]

On Thu, 9 Oct 2014, Peter Wang wrote:

> On Thu, 9 Oct 2014 13:32:58 +1100 (EST), Julien Fischer <jfischer at opturion.com> wrote:
>>>>
>>>> impure_init needs to be (1) a function and (2) defined in the public interface,
>>>> since users require a way of initialising semaphore valued mutables.
>>>
>>> Why?  If the rest of the semaphore library was pure I/O code was this ever
>>> used outside of the standard library?
>>
>> The version history is a bit confused because these modules were
>> originally moved out of extras directory back when we were using CVS.
>> It looks as though the purity of semaphore.new was originally incorrect
>> and later corrected by being made impure.  (At some point we changed
>> all the 'new' predicates to 'init'.)
>>
>>> I'm not if we do need to provide such a function, at least the
>>> deprecation will allow people to continue to use this and we can find
>>> out if it causes a problem for anyone.
>>
>> I think mutable initialisation is a legitmate reason for this to be
>> provided.
>
> Bug #223

Ok, so the _function_ impure_init still needs to be public

For consistency, the function mvar.impure_init should be introduced.

Cheers,
Julien.