(original) (raw)

On 01May2019 1305, Gerald Wiltse wrote:
\> That does help, thanks so much for the fast response! I now see how
\> it's executed on Azure pipelines also, so that should be enough for me
\> to reverse engineer the rest.
\>
\> Followup questions if you have time:
\> 1\. It seems that directory doesn't exist in the 3.6 branches/tags, so it
\> must have been done a different way back then. Do you think it's
\> back-portable to 3.6?

Yep, it should work with 3.6\. There were scripts under Tools/msi back
then that did the same thing, but with the Store app I rewrote it to be
more flexible.

\> 2\. One of our objectives is to effectively create a nearly identical
\> embeddable package for linux. Do you think the layout code is generic
\> enough to be modified to create a linux equivalent?

Probably not, it's under the "PC" directory because it is entirely
CPython specific.

I'd love to be able to have a similar embeddable distro for Linux (and
macOS), and I'll be investing some of my time in working towards this,
but it's far more complicated than simply laying out the right files. If
you're interested in helping with this, then we're in the right place to
discuss it, but we'll likely have to move to the issue tracker at some
point once we figure out who else wants/needs to be involved.

\> Obviously, there are many alternative methods and documents for creating
\> "embeddable" pythons on linux, but the windows package has all the
\> characteristics we want, and of course we'd like our Windows and Linux
\> process build and packaging processes to be as similar as possible, so
\> it seems attractive to try to port this layout script to linux.

"make install" already does the same thing, but the install location
matters at build time, so it's a little more complex to make it work as
an embedded runtime. I'm sure there are people who know how to make it
work though.

Cheers,
Steve