[U-Boot] [PATCH] test/py: Fix pytest4 deprecation warnings

[Tom Rini]

On Wed, Mar 13, 2019 at 09:01:53PM -0400, Tom Rini wrote:
> On Thu, Mar 14, 2019 at 01:20:09AM +0100, Marek Vasut wrote:
> > On 3/13/19 8:42 PM, Tom Rini wrote:
> > > On Wed, Mar 13, 2019 at 07:23:15PM +0100, Marek Vasut wrote:
> > >> On 3/13/19 5:01 PM, Tom Rini wrote:
> > >>> On Wed, Mar 13, 2019 at 12:30:59PM +0100, Marek Vasut wrote:
> > >>>> On 3/13/19 12:29 PM, Tom Rini wrote:
> > >>>>> On Wed, Mar 13, 2019 at 12:27:38PM +0100, Marek Vasut wrote:
> > >>>>>> On 3/13/19 12:25 PM, Tom Rini wrote:
> > >>>>>>> On Wed, Mar 13, 2019 at 12:20:49PM +0100, Marek Vasut wrote:
> > >>>>>>>> On 3/13/19 12:19 PM, Tom Rini wrote:
> > >>>>>>>>> On Wed, Mar 13, 2019 at 05:08:14AM +0100, Marek Vasut wrote:
> > >>>>>>>>>
> > >>>>>>>>>> Fix the following spit from pytest:
> > >>>>>>>>>>
> > >>>>>>>>>> u-boot/test/py/conftest.py:438: RemovedInPytest4Warning: MarkInfo objects are deprecated as they contain merged marks which are hard to deal with correctly.
> > >>>>>>>>>>   Please use node.get_closest_marker(name) or node.iter_markers(name).
> > >>>>>>>>>>   Docs: https://docs.pytest.org/en/latest/mark.html#updating-code
> > >>>>>>>>>>     for board in mark.args:
> > >>>>>>>>>>
> > >>>>>>>>>> In both cases, the later suggestion is applicable.
> > >>>>>>>>>>
> > >>>>>>>>>> Signed-off-by: Marek Vasut <marek.vasut+renesas at gmail.com>
> > >>>>>>>>>> Cc: Igor Opaniuk <igor.opaniuk at linaro.org>
> > >>>>>>>>>> Cc: Tom Rini <trini at konsulko.com>
> > >>>>>>>>>> Cc: Simon Glass <sjg at chromium.org>
> > >>>>>>>>>
> > >>>>>>>>> Deferred, for now we don't support newer pytest than 2.8.7 and you'll
> > >>>>>>>>> need to use virtualenv to set that up if needed.  There is not, AFAICT,
> > >>>>>>>>> a way to support both versions.
> > >>>>>>>>
> > >>>>>>>> That's what's in debian testing though, so maybe we need to support it
> > >>>>>>>> somehow.
> > >>>>>>>
> > >>>>>>> Yes, I'm _very_ frustrated at the speed at which pytest went from "this
> > >>>>>>> is the API" to "this API is deprecated" to "this API doesn't work and
> > >>>>>>> here's the new, incompatible API".  Debian/testing needs to use
> > >>>>>>> virtualenv to setup a python area with older pytest installed, just like
> > >>>>>>> we do in .travis.yml.
> > >>>>>>
> > >>>>>> Can't we rather have people use the new APIs and virtualenv new python?
> > >>>>>
> > >>>>> Not as easily, no.  Debian/testing may have something much newer but
> > >>>>> Debian/stable doesn't, and I don't know what Ubuntu/18.04 has off-hand
> > >>>>> but it's probably inbetween and so on.
> > >>>>
> > >>>> While I'm not a python expert, shouldn't virtualenv help with that ?
> > >>>
> > >>> Yes, and breaking old setups is usually frowned upon and making new
> > >>> setups conform to the existing ways is how things are usually done.
> > >>
> > >> If you use venv with old setup, won't that give you the new python you
> > >> need ?
> > > 
> > > No, you don't need to.  Travis is special in that it's based on Ubuntu
> > > 14.04 (!!!!) and so we needed to use pip there to setup anything, and
> > > have for forever.  That in turn lead to us hitting this problem a while
> > > back, when "pip install pytest" first gave us something where the old
> > > behavior became a fatal error.  That leads to installing the last
> > > version before pytest starts to complain about changing APIs.  Normally
> > > old distributions however ship with 2.8.7 anyways and don't need
> > > virtaulenv.
> > 
> > I don't think I get your point here. Sure, old distros might need to
> > change and start using virtualenv because the software is just too old.
> > We cannot support all kinds of old stuff. If the API we're using is
> > getting deprecated, let's just switch to the new one and ask the users
> > of old software to upgrade (?).
> 
> My point is that "pin to a newer pytest version" is not something I want
> right now.  It will break existing setups and provide nothing in return.
> There's not some new feature of pytest we're missing out on.  My take
> away from all of this is that we need to move the whole thing into being
> wrapped up, for everyone, as we cannot expect random community python
> modules to remove an API in an extremely quick fashion.  If you're
> motivated enough over this to go off and do that, yes, sure.  But I will
> not take a patch this patch as-is, as it breaks travis-ci.  I will not
> take a v2 of this patch that is the above plus pinning to a new pytest
> as that's just going to push the problem from "Debian/Buster and others
> people need to continue to setup virtualenv" to "Ubuntu 16.04 and other
> people now need to setup virtualenv".  That's just pushing the problem
> around and not making anything better.

I might not have been clear enough here, so let me elaborate a little
more.  We do not follow now-normal conventions for using python modules.
We need to change things such that 'make tests' does a 'pip -r
requirements.txt' so that we use pip to install everything, and we need
to setup a virtualenv to do that in.  At that point we'll be following
normal conventions and best practices (more closely anyhow, ugh, test.py
is another python2 script) and can easily move from pytest 2.8.7 to
whatever-later-version because we can do it atomically and not break
anyone.  As a bonus we'll have better control over our QA environment
which is a good thing.  Thanks!

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190317/f837486d/attachment.sig>