[petsc-dev] Do we still need to support python 2.4?

[Barry Smith]

> On Oct 13, 2015, at 5:43 PM, Matthew Knepley <knepley at gmail.com> wrote:
> 
> On Tue, Oct 13, 2015 at 5:38 PM, Satish Balay <balay at mcs.anl.gov> wrote:
> Sorry - I was looking at the wrong date - and ended up captured e-mail
> from sept 2013.
> 
> Here are the numbers for e-mail starting aug-2014.
> 
>      10 2.4
>     162 2.6
>     310 2.7
> 
> I would trade 2% of support requests for parallel configure.

   You are assuming that the number of configure.log sent to us is uniformly randomly distributed among all the python versions. I might assume that configure just works much better for python 2.4 users so that they rarely need to send configure.log :-).

   But, based on the data I have to agree with you. I am fine with moving the python required version up to 2.6  but note that it is important that configure print a very helpful message when the users python is too old.  For example improve the current message by printing

  print '*** You must have Python2 version 2.6 or higher to run ./configure        *****'

                        Try python2.7 ./configure or python2.6 ./configure 


  print '*          Python is easy to install for end users or sys-admin.              *'
  print '*                  http://www.python.org/download/                            *'
  print '*                                                                             *'
  print '*           You CANNOT configure PETSc without Python                         *'
  print '*   http://www.mcs.anl.gov/petsc/documentation/installation.html     *'
  print '*******************************************************************************'

  Barry



> 
>   Matt
>  
> Total: 482
> 
> Satish
> 
> On Tue, 13 Oct 2015, Satish Balay wrote:
> 
> > For emails starting from sept 2014 - I have 865 configure.log files
> > [with python version listed].
> >
> > Here is the count for various versions:
> >
> >     28  2.4
> >     284 2.6
> >     553 2.7
> >
> > Note: most of our nightlybuild machines are using ubuntu 12.04 - which defaults to python-2.6
> >
> > Satish
> >
> >
> > On Tue, 13 Oct 2015, Barry Smith wrote:
> >
> > >
> > >   Satish,
> > >
> > >   Can you check all the configure.log files we have received in that last year for the python version? If a vanishing small number are below 2.7 we can make 2.7 the minimum requirement. Or 2.6   With your pine mail you should be able to write a python3 script in just a few minutes to check this.
> > >
> > >
> > >   Barry
> > > > On Oct 13, 2015, at 2:36 PM, Satish Balay <balay at mcs.anl.gov> wrote:
> > > >
> > > > Previously - I've used the currently supported RHEL as a base case.
> > > >
> > > > Since RHEL5 is still currently supported -and it defaults to
> > > > python-2.4 we've used python-2.4 as the minimum for petsc configure.
> > > >
> > > > However its 8 years old - [and now RHEL had much longer lifecycle 11+
> > > > years - in extended-support mode]
> > > >
> > > > I'm not sure how many users still have this old version [or other
> > > > software stack that defaults to python-2.4]. Perhaps its ok to upgrade
> > > > the requirements in petsc master [I'm not completely sure about it.\
> > > >
> > > > And RHEL-6 defaults to python-2.6. [so when we update the minimum
> > > > version - I would go for that]
> > > >
> > > > Satish
> > > >
> > > >
> > > > On Tue, 13 Oct 2015, Barry Smith wrote:
> > > >
> > > >>
> > > >>   This depends on what (prehistoric) distributions of Linux are still commonly used that only have 2.4. In our experience many users still use very old distributions and asking them to "upgrade" is not likely to be successful since the users don't control the machines they use.
> > > >>
> > > >>  This is why we kept 2.4 I think we need to keep 2.4 support
> > > >>
> > > >>  Barry
> > > >>
> > > >>> On Oct 13, 2015, at 2:19 PM, Tobin Isaac <tisaac at ices.utexas.edu> wrote:
> > > >>>
> > > >>>
> > > >>> Honest question.  I've got some changes in buildsystem
> > > >>> (tisaac/buildsystem-feature-parallel) that heavily use the
> > > >>> with-statement context manager introduced in python 2.5.
> > > >>>
> > > >>> Cheers,
> > > >>> Toby
> > > >>>
> > > >>
> > > >>
> > > >
> > >
> > >
> >
> >
> 
> 
> 
> 
> -- 
> What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.
> -- Norbert Wiener