ISLocalToGlobalMapping: free globals and manage correctly the negative indices

First attempt to allow changing the block size of a matrix (and any associated l2gmap)

minor fixes to manual pages due to bad formatting; a few Tao routines that can use const for arrays

ViennaCL: Fixed parallel runs with block-Jacobi.

Now it is possible to run ViennaCL with multiple MPI ranks again.

make type of PetscInt argument explicit so Fortran compiler does not error

Reported-by: nightly tests

Merge branch 'maint'

fix again..

fix breakage with 64bit-indices

Merge branch 'knepley/fix-doc-f90-vec'

* knepley/fix-doc-f90-vec:
Doc: Improved F90 array handling docs Reported-by: William F Mitchell <william.mitchell@nist.gov>

Merge branch 'barry/fix-dmdacreate'

Merge branch 'barry/use-fortran-kind-consistently' into barry/fix-dmdacreate

Need to resolve conflicts with .F90 code reformating

Doc: Improved F90 array handling docs
Reported-by: William F Mitchell <william.mitchell@nist.gov>

Prevent using sowing on VecMin (optional argument requires special handling in Fortran, just like VecMax).

Fortran updates

Updated all F90 examples to use free form with full error checking
Update all Fortran examples to check error code for PetscInitialize() call

Time: 2 hours

Cannot use return in Fortran main program so use stop instead

Reported-by: nightly tests

added some more printdot to testexamples_fortran rules since that is timing out with travis

Reported-by: travis

Introduce PETSC_REAL_KIND and update all .F90 examples to free form with error checking and to use it

The reason for PETSC_REAL_KIND is there is no way to assign constants (say .1) in Fortran so tha

Introduce PETSC_REAL_KIND and update all .F90 examples to free form with error checking and to use it

The reason for PETSC_REAL_KIND is there is no way to assign constants (say .1) in Fortran so that they have the full precision of the left hand side variable. And if you mark them with D0 then compile in single precision you get a warning about truncating precision

Time: 6 hour
Reported-by: Todd Munson <tmunson@mcs.anl.gov>
Thanks-to: Blaise A Bourdin <bourdin@lsu.edu>, Lisandro Dalcin <dalcinl@gmail.com>

show more ...

Merge branch 'maint'

VecShift() function pointer was missing shift argument.

Time: .1 hours
Reported-by: Jose E. Roman <jroman@dsic.upv.es>

Merge remote-tracking branch 'origin/maint'

Merged in dalcinl/fix-dmplex-global-offset (pull request #541)

DMPlex: Fix global end of range

Merge branch 'maint'

bugfix for VecTDotBegin
Contributed by: Jose E. Roman <jroman@dsic.upv.es>

bugfix for VecDotBegin()
Reported-by: Jose E. Roman <jroman@dsic.upv.es>

fix vec-ex28.c for complex build
Reported-by: Jose E. Roman <jroman@dsic.upv.es>