[petsc-dev] related to compiling your source code

Satish Balay balay at mcs.anl.gov
Wed Apr 15 10:29:27 CDT 2015


On Wed, 15 Apr 2015, Mark Adams wrote:

> > >
> > > If everyone is using ~/.petscrc then I can suck it up but if nobody is
> > then this is a red hearing and we should just do it.
> >
> >   You are avoiding the question. If we get rid of .petscrc how can we
> > possible not get rid of PETSC_OPTIONS since it has the exact same possible
> > problems (and we have plenty of previous experience with outdated
> > environmental variables)
> 
> 
> Barry, I have answered this question.  Let me try to spell this out
> explicitly.
> 
> 1) Does anyone need ~.petscrc other than Satish?  If not then kill.
> (Satish, or another expert user, is a big boy and he can take care of
> himself. eg, he can use PETSC_OPTIONS.)
> 
> 2) If so then is PETSC_OPTIONS the only/best alternative?
> 
> 3) If so, then what is cost of moving users to it?  (Data on number of
> users needed, so far 0/2)
> 
> 4) Given the cost of the number of users to switch and our cost factor for
> switching users (ie, epsilon), is it a better solution long term?
> 
> 5) I think it is: again, the problems that I have seen over the years is
> *not* someone forgetting that they *intentionally* once put a .petscrc file
> in their home directory but that the use their home directory as a scratch
> space and have a .petscrc file there by accident.  .bashrc files at least
> have to be explicitly edited an are not used as a scratch pad.  Home
> directories are use as a scratch pad by many users.

This doesn't make sense to me. 2 issues you raise:

1 *intentionally* create a .petscrc.

2. use $HOME as scratch and stage files from other machine or e-mail
[an extremely bad practice isrrespective of using petsc or not] - but
they'll *never* stage .bashrc from the other machine.

With either of these claims the said user is using .petscrc [either on
current machine or the other machine from where he/she staged files
over] so your claim of 'Satish is the only .petscrc' user is false.

Note: I don't believe the exception to .bashrc you claim. I know folks
usually copy their prefered .bashrc (or copy/paste whole chunks) to
any new machine accounts they get - instead of starting to configure
such a thing from scratch.

Anyway - I think Jed's suggestion of adding the 'source' of any
used/unused options to -options_table/-log_summary is a good
compormise..

Satish

> 
> At this point is is not clear that we even need to get past (1).
> 
> Mark
> 




More information about the petsc-dev mailing list