On 02/25/2014 04:23 PM, Dave Airlie wrote:
On Wed, Feb 26, 2014 at 6:36 AM, Dylan Baker <[email protected]> wrote:
On Tuesday, February 25, 2014 02:01:56 PM Ken Phillis Jr wrote:
On Mon, Feb 24, 2014 at 9:10 PM, Matt Turner <[email protected]> wrote:
On Mon, Feb 24, 2014 at 11:47 AM, Ilia Mirkin <[email protected]>
wrote:
On Mon, Feb 24, 2014 at 2:42 PM, Dylan Baker <[email protected]>
wrote:
This series makes the transition from python2 to python3. In general
Out of curiousity, what's the motivation for this? I've been able to
avoid python3 quite nicely thus far, all it seems to do is break
compatibility with existing setups... Admittedly piglit is targeted at
developers, who probably have the latest and greatest stuff, so
perhaps not a huge issue.
Python 2.7 support sucks. Piglit with concurrency was simply broken
with Python 2.7.4 and 2.7.5 because of some regression related to
threading. Then piglit was broken again after the timeouts code was
committed in November, until it was reverted, as far as we can tell
because of a Python bug. And Python 3.3 just has the necessary support
for doing this without rolling it ourselves.
Basically, we're tired of relying on old and dead versions of python
which are ill maintained and break often.
I agree with the forced update to Python 3, but I would suggest making
some tweaks to keep this working on a lot of the long-term desktop
environments because it is better to have a stable software base when
writing tests and improving the actual driver.
Debian: https://packages.debian.org/wheezy/python3 - The minimum
supported version of the releases is Python3 version 3.1.3
Ubuntu: http://packages.ubuntu.com/lucid/python3 - Minimum Python3
version of 3.1.2 ( on Ubuntu 10.04 LTS ), but the current long term
release ( 12.04 LTS ) features python 3.2.3.
Gentoo: This always tends to carry the latest python release so this
probably is not going to be an issue.
FreeBSD: FreeBSD 9 includes python version 3.2.3 ( or similar ) in a
package so this should work out well.
Red Hat based distributions: These appear to only have Python 2.7
based on the package searches I can find. I may be wrong, but it would
be a good idea to at least verify this.
on python3<3.2: Python as a project doesn't support python older than 3.2, and
3.1 lacks features we're already using (ie: argparse). It would be a step
backwards to go to python 3.1.
on python 3<3.3: Python 3.2 has 99% feature parity with 2.7, so while it would
be a step toward the future, many of the reasons for going to python3 are new
features that will be useful or allow us to replace our own hand rolled code
with code from the standard library. Thread timeouts is the most obvious
example. If we need to postpone 3.3 support and move to 3.2 that is an option,
although IMHO it would be better to just go all the way.
on LTS/stable releases: Piglit is a testing framework for *upstream
development* of graphics drivers. You need up-to-date libdrm, linux, llvm, etc
to even build mesa. I don't know of any developers trying to do upstream
development on EL, Debian<testing (jessie has 3.3.4), or Ubuntu<latest (saucy
has 3.3.x); though I might be wrong, and they can yell at me for making stupid
assumptions!, because the point of LTS/Stable releases is that *nothing ever
changes*. I think the right solution for LTS/stable releases is to just tag
the last 2.6+argparse compatible release, and point people who don't have 3.3
support to that tag.
We do this quite often, RHEL6 is on Mesa 9.2 and we've updated it from
Mesa 7.11 on a regular schedule,
and we do run piglit against it, so yes a false assumption.
It sounded a little bogus to me even without your data. :) Would a
python 3.3 requirement for testing be a problem?
Dave.
_______________________________________________
Piglit mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/piglit
_______________________________________________
Piglit mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/piglit
