[petsc-dev] petsc patch number

[Kong, Fande]

On Wed, Apr 19, 2017 at 10:27 AM, Satish Balay <balay at mcs.anl.gov> wrote:

>
> On Wed, 19 Apr 2017, Kong, Fande wrote:
>
> > On Wed, Apr 19, 2017 at 9:59 AM, Satish Balay <balay at mcs.anl.gov> wrote:
> >
> > > Why do you need to distinguish (between petsc maint snapshots) if you
> > > are using the fixed superlu_dist version with both?
> > >
> >
> > The old one has been complied several months ago, and we do not want to
> > recompile it,
>
> What do you mean by old? PETSc-3.6.4, or PETSc-3.7.5? [or both?]
>

PETSc-3.7.5


>
> > and most of tests stay with this version.
>
> Which version?
>

3.7.5


> > The current patched one is used for a few tests only.
>
> How would you do that? Is that a 3rd test run in addition to the above
> two?


Yes, this is a 3rd test run.


> If so why do you need version info here [why not run the third
> test with the extra configure option I mentioned with petsc-3.7.5 or
> whatever 'maint' snapshot your test scripts can grab?[
>

> Or why not fix your test scripts to grab URLs - and not hardcode version
> info?
>
> >  It is not that trivial for me to update all user side applications
> > with petsc. We therefore avoid updating petsc frequently.
>
> You appear to have basic misunderstanding about petsc patch
> releases. As Jed mentioned the API is not supporsed to change with
> patch releases so you shouldn't have to 'update user side
> applications'.
>

I know that APIs do not change, but this does not mean that numerical
behaviors  will stay the same.


>
> Anyway - I guess I don't really understand the problem you are trying
> to solve - and how a different petsc 'version' helps with it.
>
> Perhaps a 3.7.6 release will help. You'll have to wait for that.
>

I have to wait for 3.7.6. When it will be released?

Fande,



>
> Satish
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20170419/1b803743/attachment.html>