[m-rev.] for review: fix bug #226 - incorrect purity of lazy.read_if_val/2

[Julien Fischer]

On Tue, Jan 8, 2013 at 12:10 PM, Paul Bone <paul at bone.id.au> wrote:
> On Tue, Jan 08, 2013 at 11:06:21AM +1100, Julien Fischer wrote:
>> For review by Paul.
>>
>> The comments in the bug report mentioned that we were awaiting some changes
>> to the deep profiler before doing this, but since there is currently
>> only a single
>> call to read_if_val/2 in the deep profiler, I assume that these have been done.
>
> No, that change never got made.  But I'm not sure that it is necessary (I've
> forgotten some of the details over the last 12 months).  Your patch probably
> makes it okay though.

zs's main concern was merge conflicts, but if there's only a single call
to read_if_val/2 I doubt there would be that many.

...

> I think this promise should be wider and the build_var_use_list should be
> impure.

An added complication being that you would need to add an impure version
of list.foldl since build_var_use_list is called using that.

> This is a change I can do later.  As it is your patch is better
> than the previous version as it at least corrects the problem in lazy.m
> Please commit it as is and open a new bug/issue for me to fix the promise.

Shall do.

Cheers,
Julien.