Hi All,
I'd like to mark current PR's for inclusion in 1.10. If there is something that you want to have in the release, please mention it here by PR #.I think new enhancement PR's should be considered for 1.11 rather than 1.10, but bug fixes will go in. There is some flexibility, of course, as there are always last minute items that come up when release contents are begin decided.
Chuck
On Mon, Apr 6, 2015 at 9:22 PM, Charles R Harris charlesr.harris@gmail.com wrote:
Hi All,
I'd like to mark current PR's for inclusion in 1.10.
Good idea. If you're going to do this, it may be helpful to create a new 1.10 milestone and keep but clean up the "1.10 blockers" milestone so there are only real blockers in there.
If there is something that you want to have in the release, please mention it here by PR #.I think new enhancement PR's should be considered for 1.11 rather than 1.10, but bug fixes will go in.
Assuming you mean "no guarantees for anything that comes in from now on", rather then "no one is allowed to merge new enhancements PRs before the release split" - makes sense.
There is some flexibility, of course, as there are always last minute items
that come up when release contents are begin decided.
I had a look through the complete set again. Of the ones that are not yet marked for 1.10, those that look important to get in are: - new "contract" function (#5488) - the whole set of numpy.ma PRs - the two numpy.distutils PRs (#4378, #5597) - rewrite of docs on indexing (#4331) - deciding on a bool indexing deprecation (#4353) - weighted covariance for corrcoef (#4960)
There are too many PRs marked as "1.10 blockers", I think the only real blockers are: - __numpy_ufunc__ PRs (#4815, #4855) - sgemv segfault workaround (#5237) - fix for alignment issue (#5656) - resolving the debate on diagonal (#5407)
Ralf
On Mon, Apr 6, 2015 at 3:01 PM, Ralf Gommers ralf.gommers@gmail.com wrote:
On Mon, Apr 6, 2015 at 9:22 PM, Charles R Harris < charlesr.harris@gmail.com> wrote:
Hi All,
I'd like to mark current PR's for inclusion in 1.10.
Good idea. If you're going to do this, it may be helpful to create a new 1.10 milestone and keep but clean up the "1.10 blockers" milestone so there are only real blockers in there.
Good idea.
If there is something that you want to have in the release, please mention it here by PR #.I think new enhancement PR's should be considered for 1.11 rather than 1.10, but bug fixes will go in.
Assuming you mean "no guarantees for anything that comes in from now on", rather then "no one is allowed to merge new enhancements PRs before the release split" - makes sense.
There is some flexibility, of course, as there are always last minute
items that come up when release contents are begin decided.
I had a look through the complete set again. Of the ones that are not yet marked for 1.10, those that look important to get in are:
Thanks for taking a look.
- new "contract" function (#5488)
- the whole set of numpy.ma PRs
- the two numpy.distutils PRs (#4378, #5597)
- rewrite of docs on indexing (#4331)
- deciding on a bool indexing deprecation (#4353)
- weighted covariance for corrcoef (#4960)
There are too many PRs marked as "1.10 blockers", I think the only real blockers are:
- __numpy_ufunc__ PRs (#4815, #4855)
- sgemv segfault workaround (#5237)
- fix for alignment issue (#5656)
- resolving the debate on diagonal (#5407)
Chuck
On Apr 6, 2015 2:01 PM, "Ralf Gommers" ralf.gommers@gmail.com wrote:
There are too many PRs marked as "1.10 blockers", I think the only real
blockers are:
- __numpy_ufunc__ PRs (#4815, #4855)
The main blocker here is figuring out how to coordinate __numpy_ufunc__ and __binop__ dispatch, e.g. PR #5748. We need to either resolve this or disable __numpy_ufunc__ for another release (which would suck).
This needs some careful attention, so it'd be great if people could take a look.
- sgemv segfault workaround (#5237)
- fix for alignment issue (#5656)
Agreed on these.
- resolving the debate on diagonal (#5407)
Not really a blocker IMHO -- if we release 1.10 with the same settings as 1.9, then no harm will be done. (I guess some docs might be slightly off.) And IMO that's the proper resolution for the moment anyway :-).
-n
On Mon, Apr 6, 2015 at 5:59 PM, Nathaniel Smith njs@pobox.com wrote:
On Apr 6, 2015 2:01 PM, "Ralf Gommers" ralf.gommers@gmail.com wrote:
There are too many PRs marked as "1.10 blockers", I think the only real
blockers are:
- __numpy_ufunc__ PRs (#4815, #4855)
The main blocker here is figuring out how to coordinate __numpy_ufunc__ and __binop__ dispatch, e.g. PR #5748. We need to either resolve this or disable __numpy_ufunc__ for another release (which would suck).
This needs some careful attention, so it'd be great if people could take a look.
- sgemv segfault workaround (#5237)
- fix for alignment issue (#5656)
I think #5316 is the alignment fix.
Agreed on these.
- resolving the debate on diagonal (#5407)
Not really a blocker IMHO -- if we release 1.10 with the same settings as 1.9, then no harm will be done. (I guess some docs might be slightly off.) And IMO that's the proper resolution for the moment anyway :-).
Asked for this to be reopened anyway, as it was closed by accident.
Chuck
participants (3)
-
Charles R Harris -
Nathaniel Smith -
Ralf Gommers