[SciPy-dev] tagging 0.7rc1 this weekend?

Jarrod Millman millman@berkeley....
Tue Dec 2 21:48:55 CST 2008


On Tue, Dec 2, 2008 at 1:52 AM, Tiziano Zito <opossumnano@gmail.com> wrote:
> #800 : numpy's test_poly1d_nan_roots is broken if scipy.linalg is imported
> #243: linalg.qr to take mode argument
> both have one or more fixes possible, a decision is needed.
>
> another IMO seriuos problem:
> #794: linalg.eig segmentation fault on unpickled arrays
> there's no fix to that one.

I don't think we are going to be able to close these tickets in time
for the 0.7.0 release, but I am not sure whether they should be
considered release blockers.  If you disagree, please let me know.

#800 -- I think we should leave scipy as is and consider this a numpy
bug and fix numpy (i.e., use asarray_chkfinite in numpy.linalg.eig)

#243 -- I think that we should make some changes to qr, but I don't
think there is a general agreement on what we should do.  There is
another thread about this, which I will look at again.

#794 -- I am not sure what to do about this, but since pickling arrays
is discouraged I don't think we should hold the release for this.  Of
course, it would be great if you could resolve this in the next few
days.

> #630: Support for generalized symmetric eigenvalue problem in SciPy
> can be closed.

Please close this ticket.  Your account (tzito) should now have
permission.  If not, please let me know.

> #593: flapack.error: ((lwork==-1) || (lwork>=MAX(1,8*n))) failed for 3rd keyword lwork
> can be closed.

Please close this ticket.

> #632: Matrix gallery aka test matrices
> should be scheduled for 0.8 (I also have some code for that).

Please feel free to take this ticket.  I for one would be interested
in seeing this in 0.8.

> #585: scipy.linalg.pinv() returns array when input is matrix
> the fix seems easy in this case, but I think a decision is needed.

I think everything should be an array (i.e., we should remove
matrices), so I would rather leave this as is.  But I am sure that
many people will disagree with me....

> #492: Problem in linalg.svd ?
> no fix. (but does the problem still exists?)

Not sure.  But I don't think this should hold up the release.

> I can work on some of these tickets if needed.

While I don't think that any of these things constitute grounds for
postponing the release candidate, I would love to see as many tickets
closed as possible.  If you have any other tickets that should be
reviewed, please let us know.

Thanks,

-- 
Jarrod Millman
Computational Infrastructure for Research Labs
10 Giannini Hall, UC Berkeley
phone: 510.643.4014
http://cirl.berkeley.edu/


More information about the Scipy-dev mailing list