[petsc-dev] possible problem with prefixed build in master

[Barry Smith]

On Jun 17, 2014, at 4:26 PM, Satish Balay <balay at mcs.anl.gov> wrote:

> It was my attempt to partly fix this issue that broke for Blaise.
> 
> I've pushed a revert for this partial/broken fix..
> 
> Wrt the issue Barry referes to - I still think 'prefix install for
> --download-package with shared libraries' is difficult. I'm not sure
> how to deal with it.

  Agreed. Maybe we should generate a useful error message in this case to prevent people wasting time trying.

   Barry

> 
> Satish
> 
> On Tue, 17 Jun 2014, Barry Smith wrote:
> 
>> 
>>  Blaise,
>> 
>>    Unfortunately our only two competent developers Jed and Satish refuse to fix this issue so unless you can fix it we are screwed.
>> 
>>    The issue is that we want only a single “make install” phase that requires the user to put in a password AFTER all the source code is compiled (and libraries linked?) but then PETSc is compiled/linked against the external libraries in original location instead of the correct final location. We cannot put them in the final location until the “make install” but by then it is too late to link PETSc libraries against them.
>> 
>>  Barry
>> 
>> On Jun 17, 2014, at 3:46 PM, Blaise A Bourdin <bourdin at lsu.edu> wrote:
>> 
>>> Hi,
>>> 
>>> there seems to be a race condition with prefix installs in petsc-dev (master) right now:
>>> external packages are built in $PETSC_DIR/$PETSC_ARCH/externalpackages, but it looks like petsc try to link with libraries in $PREFIX/lib which are not installed yes. Is this fixable before the release?
>>> 
>>> Blaise
>>> 
>>> 
>>> <configure.log.zip><make.log.zip>
>>> -- 
>>> Department of Mathematics and Center for Computation & Technology
>>> Louisiana State University, Baton Rouge, LA 70803, USA
>>> Tel. +1 (225) 578 1612, Fax  +1 (225) 578 4276 http://www.math.lsu.edu/~bourdin
>>> 
>>> 
>>> 
>>> 
>>> 
>>> 
>>> 
>> 
>>