<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0//EN" "http://www.w3.org/TR/REC-html40/strict.dtd"> <html><head><meta name="qrichtext" content="1" /><style type="text/css"> p, li { white-space: pre-wrap; } </style></head><body style=" font-family:'DejaVu Sans'; font-size:9pt; font-weight:400; font-style:normal;"> On Wednesday, February 26, 2014 10:23:31 Dave Airlie wrote: > On Wed, Feb 26, 2014 at 6:36 AM, Dylan Baker <baker.dylan.c@gmail.com> 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 <mattst88@gmail.com> wrote: > >> > On Mon, Feb 24, 2014 at 11:47 AM, Ilia Mirkin <imirkin@alum.mit.edu> > > > > wrote: > >> >> On Mon, Feb 24, 2014 at 2:42 PM, Dylan Baker <baker.dylan.c@gmail.com> > > > > 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. > > Dave. Sorry Dave, somewhere I'd gotten the impression that you guys did most of your work on Fedora, I stand corrected -Dylan</body></html>