@Raul: I was trying to say that there is no rush to get this into 9.5.
We will release what we have for 9.5 and finish your rewrite in 9.6.
We always want to give new code a couple of weeks in beta before we make
a final. The changes I have made in the past month are tiny. (It might
look like I've changed a lot, but it's all QP stuff that is not being
announced).
Since your changes seem to have the potential to break something, I
think it would be better to back them out of 9.5 and put them into 9.6.
Let me know if that is a hardship in any way.
I was planning to call for a new beta today. It will be the final
beta. The only new feature is u&.(m&{) . we can release 9.05.01 at the
end of the year and get your code into 9.06-beta1.
Please tell me when you have backed your change out of 9.5. Or, if you
feel confident that your change is solid, tell me that. I will then
call for the final beta.
hhr
On 12/4/2023 10:12 AM, Raul Miller wrote:
No, not only a private branch. I took your message
https://www.jsoftware.com/pipermail/source/2023-December/001628.html
as an indication that I should push what I had working to the main branch.
That said, I see what happened. Though I'm baffled how it did not fail
on my machine.
mpn_com is supposed to implement a bitwise negation of a span of gmp
limbs. It's mpn_neg which is supposed to implement the two's
complement of that span.
I misremembered these details and implemented in my jmpn_com the logic
of mpn_neg. That should never have worked in any examples of negating
an extended integer I can think of.
And yet it repeatedly passed tests on my machine. I don't understand
how that was possible. I guess I'm going to spend some time looking at
intermediate results before I go any further with this project.
Thanks,
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
