On Thu, Feb 27, 2020 at 12:37:46PM +0100, Jeremie Courreges-Anglas wrote: > On Thu, Feb 27 2020, Martin Reindl <[email protected]> wrote: > > On Wed, Feb 26, 2020 at 11:17:05AM +0100, Jeremie Courreges-Anglas wrote: > >> On Sun, Feb 23 2020, Martin Reindl <[email protected]> wrote: > >> > Am 21.02.20 um 15:27 schrieb Bjorn Ketelaars: > >> >> Enclosed diff updates py-pandas to 1.0.1. Changes: > >> >> https://pandas.pydata.org/docs/whatsnew/v1.0.0.html and > >> >> https://pandas.pydata.org/docs/whatsnew/v1.0.1.html. Noticeable change > >> >> is that python2 support has been dropped. > >> >> > >> >> This diff also touches net/py-siphon and databases/py-influxdb. Drop > >> >> python2 support by: > >> >> - setting FLAVOR=python3 and bumping REVISION in Makefile > >> >> - adding conflict- and pkgpath- markers to PLIST > >> >> - adding quirk to devel/quirks > >> >> > >> >> Other consumers have been left alone as these are already python3 only. > >> >> > >> >> Testing: > >> >> - 'make test' still fails some test > >> >> - build tested consumers > >> >> - run tested on amd64 with some work related scripts > >> >> > >> >> Comments/OK? > >> > > >> > Thanks for the update, Bjorn. > >> > > >> > Tests quickly segfault here on amd64 in pandas/tests/test_sorting.py. > >> > Any idea why? > >> > > >> > On the other hand with py-siphon (I will update TDEPs at a later point): > >> > 7 failed, 234 passed, 64 warnings > >> > > >> > But no failures are directly related to pandas - so OK for the siphon > >> > part. > >> > > >> > Also, a compile test on macppc and/or sparc64 would be great if anyone > >> > can spare the CPU cycles. > >> > >> make package succeeds on sparc64, make test summary: > >> > >> = 460 failed, 59252 passed, 3480 skipped, 539 xfailed, 7 xpassed, 84 > >> warnings, 153 error in 26399.67 seconds = > >> > >> make build and test log: > >> > >> https://autogeree.net/~jca/tmp/py3-pandas.log.gz > > > > With more mem, tests here pass in a similiar fashion on amd64. Very happy > > with > > py-pandas moving forward - OK martin > > Looks good to me too, but: > - quirks need to be refreshed > - math/mlpack still references an unflavored math/py-pandas
So, the patch needs to fix that as well ;)
