On 10/03/2021 20:57, Mats Wichmann wrote:
On 3/10/21 11:56 AM, Thomas Jollans wrote:
On 10/03/2021 18:42, Marco Sulla wrote:
On Wed, 10 Mar 2021 at 16:45, Thomas Jollans <t...@tjol.eu> wrote:
Why are you doing this?
If all you want is for it to be possible to install the package from
source on a system that can't use the C part, you could just declare
your extension modules optional
Because I want to provide (at least) two wheels: a wheel for linux
users with the C extension compiled and a generic wheel in pure python
as a fallback for any other architecture.
What's wrong with sdist as a fallback rather than a wheel?
That has the added benefit of people on other architectures have the
opportunity to use the extension module if they have a compiler and
the necessary libraries and headers installed...
Doesn't that mean nasty failures for those don't have the correct
build setup (like almost every Windows user on the planet)? This
isn't a snide question, I'm actually interested in solving roughly the
same problem as the OP.
I believe that this is pretty much exactly the problem that the
"optional" flag for extensions solves.
docs: "specifies that a build failure in the extension should not abort
the build process, but simply skip the extension."
I would assume this refers to *any* build failure (including a missing
compiler), so people without a proper build system might get error
messages, but should still get a working package (assuming of course the
C extensions is *actually* optional, like an accelerator module).
But I don't have experience with this myself so take what I say with a
grain of salt.
- Thomas
--
https://mail.python.org/mailman/listinfo/python-list
