DevHeads.net

python-kitchen long-standing build failures

Hey Ralph and Fedora Infra,

Could you please fix python-kitchen long standing build failures or
alternatively stop using it in infra tools and retire it?

The package is a Fedora specific downstream==upstream almost dead project with
utility methods for anything. Most of them are either one-liners or magical
attempts to remove the unicode/bytes separation from Python 3 and go back to
Python 2 where everything "just works".

The failures blocked the Python 3.7 upgrade in Fedora and now they block the 3.8
upgrade, it will block the 3.9 upgrade, etc.

I consider the kitchen unicode/str methods dangerous, so I'd prefer if we just
stop using it, however if you just fix the failing tests, it would work.

You can probably also decide to skip the tests temporarily forever and be done
with this, but the current situation is really frustrating for us.

See <a href="https://github.com/fedora-infra/bodhi/pull/3094" title="https://github.com/fedora-infra/bodhi/pull/3094">https://github.com/fedora-infra/bodhi/pull/3094</a> for reference how to drop
kitchen.

The FTBFS bug: <a href="https://bugzilla.redhat.com/show_bug.cgi?id=1605741" title="https://bugzilla.redhat.com/show_bug.cgi?id=1605741">https://bugzilla.redhat.com/show_bug.cgi?id=1605741</a>

Thank you,

PS I don't want to run nonresponsive maintainer policy for infra-sig :(

Comments

Re: python-kitchen long-standing build failures

By Randy Barlow at 05/14/2019 - 09:47

On Tue, 2019-05-14 at 14:50 +0200, Miro HronĨok wrote:
Kitchen takes the wrong approach to string encoding - I recommend that
we stop using it and retire it. It can be a dangerous library if its
string encoding stuff is used for anything security related, and it's
still bad otherwise because it hides all encoding exceptions from you,
which means you will get corrupted data in your applications.

Re: python-kitchen long-standing build failures

By Kevin Fenzi at 05/14/2019 - 12:27

On 5/14/19 6:47 AM, Randy Barlow wrote:
I have no horse in this race, but if we retire python-kitchen, someone
needs to go though the things that now depend on it and remove that
dependency. ;(

I guess short term I could make it build (I got it just working mostly,
but failing a test), but that doesn't help too long with python2
retirement coming soon. Might give us a little breathing room though?

kevin

Re: python-kitchen long-standing build failures

By =?UTF-8?B?TWlyb... at 05/14/2019 - 12:52

On 14. 05. 19 18:27, Kevin Fenzi wrote:
Making it build is my primary concern.

Stopping using it is my secondary unsolicited advice.

While the Python 2 retirement will certainly mess with things, the issue I have
is with kitchen usage on Python 3 (where unicode/bytes is done right, but
kitchen makes it python2ish again).