[petsc-dev] problem with sub autoconf builds using badly set environmental variables

[Satish Balay]

Looks like current configure already had this warning.

>>>
      if 'CC' in os.environ and os.environ['CC'] == self.framework.argDB['CC']:
        self.logPrintBox('\n*****WARNING: Using C compiler '+self.framework.argDB['CC']+' from environmental variable CC****\nAre you sure this is what you want? If not, unset that environmental variable and run configure again')
<<<

I think we force compilers on command line to most packages - so env
should never get used. [if they do - then we should supply more
options to that package-configure from petsc-configure.

MPICH configure needs 2 args for fortran [one for f77 and the other
for f90] - and this old petsc provided only one -[so perhaps mpich
configure was guessing the other one - hence the incompatibility].

Regarding the ENV stuff - this old PETSc was passing args to MPICH
configure via env variables [which we rectified in subsequent petsc
releases]. So the primary issue here is the usage of old petsc
version.

Satish

On Mon, 11 Jan 2010, Barry Smith wrote:

> 
>   If some imbecile has, say FC, set to something and then runs PETSc's
> ./configure using --with-fc=somethingelse then autoconf ./configures started
> by PETSc (for example --download-mpich)  will use the environmental variable
> instead and thus have incompatible compilers and a mess impossible to debug.
> 
>   Satish, can we have PETSc's config/configure.py STOP if any of the those
> dangerous variables are set and tell the user to get rid of them before
> running ./configure to prevent this problem and our wasted time on
> petsc-maint's?
> 
>   Thanks
> 
>    Barry
> 
> 
> 
> Begin forwarded message:
> 
> > From: "Zhao Y." <yz1c08 at soton.ac.uk>
> > Date: January 11, 2010 2:32:21 PM CST
> > To: petsc-maint Maintenance <petsc-maint at mcs.anl.gov>
> > Subject: [petsc-maint #40116] Problem with installation
> > Reply-To: petsc-maint at mcs.anl.gov, "Zhao Y." <yz1c08 at soton.ac.uk>
> > 
> > Hi, the log file is attached.Please help me figure the problem out.
> > 
> > Yours sincerely
> > Yu Zhao