This might be an option. I asked the user what is the reason the installation was done.
However I already experienced some problems with .whl files installed on different platforms. This is one of the reasons we are preparing our wheel files inside the CI container based on debian, because when they were prepared on DockerHub, they could not be installed on Debian. So I think there will be systems where sdist is used automatically because the host system is not as close to what we use when build wheels. J. On Sat, Dec 12, 2020 at 10:25 AM Ash Berlin-Taylor <a...@apache.org> wrote: > I wonder if we should not bother releasing the sdist at all? Since our > wheel is OS agnostic anyway, and wheel is much faster to install, I see > little point in continuing to release the sdist. > > We should have the artefacts become the source (`git archive`) and the > `wheel` only I think. > > -ash > > On Sat, 12 Dec, 2020 at 07:48, Jarek Potiuk <jarek.pot...@polidea.com> > wrote: > > Unfortunately, -1 from me. > > Here is the issue: https://github.com/apache/airflow/issues/13027 > > We miss a schema file in sdist package. That causes airflow commands to > fail if Airflow is installed using sdist package (.targz) rather than .whl. > We've tested everything with .whl files installation and everything works > fine, but with the .sdist package we have a problem. > > Learning for the future: we currently optimized away testing for .sdist > packages and while we actually test installations with sdist package, we > only run "airflow" in images that were build from .whl packages :( (and the > discussion at this time was "If it runs with .whl. it will also run with > sdist as soon as we see that we can install it"). Turns out - that it's not > the case. > > I will shortly add some more tests to at least run at least some smoke > tests where we install everything from sdist packages as well. > > I already created a PR fixing that > https://github.com/apache/airflow/pull/13031 > > I also created a new milestone > https://github.com/apache/airflow/milestone/24 "Airflow 2.0.0rc2" where I > added a few other candidates (mostly doc and dev tools) which we can decide > to cherry-pick for RC2. > > I propose (Kaxil/Ash to confirm as release managers) that we should > continue running the testing with rc1 (if someone actually has some time to > spare during the weekend) and I believe that we release RC2 early next > week. > > J. > > > > > > > On Thu, Dec 10, 2020 at 9:45 PM Daniel Standish <dpstand...@gmail.com> > wrote: > >> Also inexplicitly got error `scheduler error (sqlite3.OperationalError) >>> database is locked`. >> >> sorry: inexplicably 🤦♂️ >> >>> > > -- > > Jarek Potiuk > Polidea <https://www.polidea.com/> | Principal Software Engineer > > M: +48 660 796 129 <+48660796129> > [image: Polidea] <https://www.polidea.com/> > > -- Jarek Potiuk Polidea <https://www.polidea.com/> | Principal Software Engineer M: +48 660 796 129 <+48660796129> [image: Polidea] <https://www.polidea.com/>
