if you mean he connection error that's a flake. The kartothek build nightly passes in the last 3 reports.
On Sun, Apr 12, 2020 at 3:57 AM Wes McKinney <wesmck...@gmail.com> wrote: > > I'm not able to reproduce the test-conda-python-3.7-kartothek-master > failure locally with docker-compose, is that a flake or real? > > On Sat, Apr 11, 2020 at 7:29 AM Krisztián Szűcs > <szucs.kriszt...@gmail.com> wrote: > > > > On Sat, Apr 11, 2020 at 12:37 PM Antoine Pitrou <anto...@python.org> wrote: > > > > > > > > > Le 11/04/2020 à 12:34, Krisztián Szűcs a écrit : > > > > - test-conda-python-3.7-turbodbc-latest: > > > > - test-conda-python-3.7-turbodbc-master: > > > > The latest is important here, because the release would break the > > > > interoperability with turbodbc. > > > > I need feedback on this from Uwe. > > > > > > For transparency, this was diagnosed and two issues were filed for > > > turbodbc: > > > https://github.com/blue-yonder/turbodbc/issues/251 > > > https://github.com/blue-yonder/turbodbc/issues/250 > > > > > > I'm not sure there's anything Arrow can do to restore compatibility. If > > > we relax array validation to avoid the validation failure that turbodbc > > > is getting, the erroneously-constructed Arrow array could lead to > > > further errors or crashes later on. > > Than we shouldn't consider the turbodbc integration tests as blocker. > > > > Thanks Antoine! > > > > > > Regards > > > > > > Antoine.