R7RS promise bugs (on corner cases) about picrin

[nyuichi]

@KeenS

Oh, tha sounds crazy... But that's a good explanation. I understood it.

[KeenS]

Hmm that's interesting. As you described in your blog, what matters seems reentrancy.

Could I ask you to send us patch? Here the code and the patch won't be so big. Thank you.

[stibear]

I tried to fix these bugs and have fixed the second bug easily.

However, to fix the first problem requires to overwrite pointers (ref. http://srfi.schemers.org/srfi-45/srfi-45.html : this uses a list as a 'box' instead of the pointer).

In our implementation, a promise is a record. I'm worried that wrapping by a list takes some merits of record types from our promise.

What do you think of this?

[KeenS]

Oh, sorry @yhara I asked you to send patch but @stibear has done the task.

[yhara]

@KeenS no problem :-)

In our implementation, a promise is a record. I'm worried that wrapping by a list takes some merits of record types from our promise.

One merit of implementing promise as record is that you can hide implementation details from user. From that point of view, wrapping box with record is better than wrapping record with box.

[nyuichi]

@yhara @KeenS @stibear

Hi, guys

Could someone tell me why the second test is expected to be ok? I still don't figure out it. Just pointing the paragraph in the spec is enough.

Thanks.

[KeenS]

@nyuichi it's sensitive, but I found this sentence.

The value of the promise must be cached (or “memoized”) so that if it is forced a second time, the previously computed value is returned.

At the first call of force, the expression is expected to return ng but in the expression, the second call of force return ok (that's the first return from force). Apparently, the first computed value is the first returned value.