[gentoo-dev] Status of Python 3.8

[gentoo-dev] Status of Python 3.8

[Michał Górny]

Hi, everyone.

I think it's time to ask once again: how much do you need Python 3.8
target to remain available in Gentoo?

Right now we support four CPython targets: 3.8 through 3.11.  3.10 is
the current default, we're going to move 3.11 from ~arch to stable soon.
We've also added the first alpha release of Python 3.12 but it's not
going to make it into PYTHON_TARGETS until the feature freeze (planned
for May 2023).

Right now 3.8 and 3.9 are both in "security" supported state upstream,
i.e. they no longer receive bugfixes except for (some of) security
backports.  Backporting more security fixes hasn't been much of
a problem, though it means we're slowly diverging from vanilla installs.

Python packages tend to continue providing support or at least working
with Python 3.8.  I suspect that if a package is missing 3.8 target,
it's more often because it wasn't tested than because it doesn't work.

So, no major reason to remove Python 3.8, except that testing packages
on four interpreters obviously takes more time than on three, and this
is starting to become noticeable.

-- 
Best regards,
Michał Górny

Re: [gentoo-dev] Status of Python 3.8

[Anna V]

On 2022-11-06 07:09, Michał Górny wrote:
> Hi, everyone.
> 
> I think it's time to ask once again: how much do you need Python 3.8
> target to remain available in Gentoo?
> 
> Right now we support four CPython targets: 3.8 through 3.11.  3.10 is
> the current default, we're going to move 3.11 from ~arch to stable soon.
> We've also added the first alpha release of Python 3.12 but it's not
> going to make it into PYTHON_TARGETS until the feature freeze (planned
> for May 2023).
> 
> Right now 3.8 and 3.9 are both in "security" supported state upstream,
> i.e. they no longer receive bugfixes except for (some of) security
> backports.  Backporting more security fixes hasn't been much of
> a problem, though it means we're slowly diverging from vanilla installs.
> 
> Python packages tend to continue providing support or at least working
> with Python 3.8.  I suspect that if a package is missing 3.8 target,
> it's more often because it wasn't tested than because it doesn't work.

Some packages fail tests on py3.8, see:
https://github.com/fastavro/fastavro/issues/558

And it's often unclear whether a package is broken on a specific Python
version or on all of them. So having lots of targets is a problem for
debugging.

> So, no major reason to remove Python 3.8, except that testing packages
> on four interpreters obviously takes more time than on three, and this
> is starting to become noticeable.