[Biopython-dev] [SPAM] Fwd: Important notice about your projects on PyPI

[Ben Fulton]

Yes, I built it from a branch. You can see that here:

https://github.com/benfulton/biopython/tree/appveyor

and the build logs are here:

https://ci.appveyor.com/project/benfulton/biopython

Where you will see that tests were run and wheels, installers, and MSI's
were built for seven versions of Python, both 32- and 64-bit versions.
(Actually, tests failed on 2.6.6 and 2.7.0...either something to do with
the shell=true issue or an inability to write files for some reason.)

I assume Jython would just be a matter of installing some software, but I'm
not sure if they support specific operating systems like XP. That would be
more a testing issue than a build issue I would guess, though.

Ben Fulton


On Thu, Oct 8, 2015 at 4:49 AM, Peter Cock <p.j.a.cock at googlemail.com>
wrote:

> On Thu, Oct 8, 2015 at 3:17 AM, Ben Fulton <ben at benfulton.net> wrote:
> > I was able to get an AppVeyor build going. It requires a top-level file
> and
> > a directory with a few configuration files in it. To get everything to
> build
> > I had to make a few changes, though:
> >
> > - Change the version from 1.65+ to 1.65, as the + isn't recognized by all
> > versions
>
> If release tools abort on the plus suffix that's usually a feature :)
> We switch to a "proper" version number during the release build
> process anyway. We could use something with the git hash but
> that isn't OK according to https://www.python.org/dev/peps/pep-0440/
>
> I suppose a fixed -dev suffix might be a good compromise?
>
> > - Add "2012Server" to the list of Windows versions in which to set
> > shell=true. (With that change I do wonder if there is any OS where shell
> > needs to be false at all)
>
> Another tweak like this? Yes please for another pull request:
>
> https://github.com/biopython/biopython/commit/780c2459cd1e76cb0a1a030de8b4cdec6dcef7b4
>
> You may be right about shell=True as a simpler option.
>
> > - Change setup.py to import setup from setuptools rather than
> > distutils.core. This is required to get bdist_wheel as a setup command.
> I'm
> > not sure if there are any issues with importing just this one function
> from
> > setuptools and the others from distutils.
>
> Not sure, Brad may have some thoughts on this?
>
> > After that I did get several wheels to build, though. If someone wants to
> > associate an Appveyor account with the Biopython Github account, I'd be
> > happy to submit a pull request with these changes.
> >
> > Ben Fulton
>
> Is this work on one of your branches?
>
> Do you see Appveyor as offering both continuous integration
> testing on Windows (yay!) and a way to build installers?
>
> We have TravisCI for Linux, potentially with Mac OS X support.
> http://testing.open-bio.org/biopython/tgrid
>
> We also run a buildbot instance to get nightly cross platform
> testing but it depends on volunteer buildslave machines - right
> now my old Windows XP box is offline for some reason, while
> the Java on another machine is too old for the latest Jython.
> http://testing.open-bio.org/biopython/tgrid
>
> Thanks,
>
> Peter
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.open-bio.org/pipermail/biopython-dev/attachments/20151008/a97ed7c0/attachment.html>