Merge branch 'balay/match-cxx-fc-with-c'

Merge branch 'shri/update-SuiteSparse'

syntax problem in BlasLapack.py

need to mangle blas/lapack symbol before testing with shared library

trigger error message if Blas/lapack libraries cannot be used with PETSc shared libraries

added downloader for fblaslapack

cleanup of duplicate code in BlasLapack.py and fblaslapack.py and f2cblaslapack.py by moving package.py

split fblaslapack.py into its own file

configure: attempt to match cxx fc compilers with the c compiler [when guessing compilers]

configure: add FC_DEFINE_FLAG to petscvariables

update to latest netcdf
fixes bug that did not allow building on recent Mac OS X

mpich: 3.1 fails to build with gcov - so revert to 3.0 [if gcov is enabled]

solaris: fix typo

openmpi: also update openmpi to version 1.6.5

mpich: update to latest release 3.1 [continue to use 3.0-patch tarball on solaris]

fixed hdf5.py so that if self.libraries.compression is None the code still runs correctly

build: split cxx [legacy] build target via SOURCECXX

Merge branch 'jed/elemental-int64'

* jed/elemental-int64:
Elemental: disable Qt5 so we don't have to pass -I for those include paths
nightlybuilds: disable elemental build when conflicting with

Merge branch 'jed/elemental-int64'

* jed/elemental-int64:
Elemental: disable Qt5 so we don't have to pass -I for those include paths
nightlybuilds: disable elemental build when conflicting with hdf5
elemental: update nightly builds. [add with-cxx-dialect=C++11 to linux - disable elsewere]
Elemental: use self.framework.getCompilerFlags()
Elemental: add support for 64-bit indices
Elemental: upgrade to development version that requires C++11
config: add with-cxx-dialect and check for C++11
Elemental: fix setting of CMAKE_CXX_FLAGS
config: write CXX_FLAGS for building C++ even when clanguage=C

show more ...

Merge branch 'barry/fix-chaco'

check for ddot_() in chaco and reject it
print download packages before downloading

update-SuiteSparse: Need to also remove CHOLMOD and UMFPACK from list of package names in config/BuildSystem/config/packages/__init__.py

update-SuiteSparse: Rempved UMFPACK.py and CHOLMOD.py, these packages
get installed now by SuiteSparse.py

f2cblaslapack: fix quad version of ila*l[rc] i*max1 functions

Fixed-by: Eloy Romero Alcalde <elroal@upvnet.upv.es>

Reported-by: Barry Smith <bsmith@mcs.anl.gov>

I stumbled across bugs in f2c

f2cblaslapack: fix quad version of ila*l[rc] i*max1 functions

Fixed-by: Eloy Romero Alcalde <elroal@upvnet.upv.es>

Reported-by: Barry Smith <bsmith@mcs.anl.gov>

I stumbled across bugs in f2cblaslapack-3.4.2.q/ for quad/__float128 precision that need fixing.

In the Lapack directory ila*lc.c ila*lr.c are single/double real/complex versions but the toclapack.sh
(or whatever generates the quad precision files) does not generate the needed ilaqlc.c ilaqlr.c and ilawlc.c
ilawlr.c files. In addition, calls to iladlc_ and iladlr_ (same for z) in the q/w source files are not
correctly mapped to the q (same for w) versions instead they are left as is (in for example qlarf.c)

extern integer iladlc_(integer *, integer *, quadreal *, integer *),
iladlr_(integer *, integer *, quadreal *, integer *);

/* Scan for the last non-zero column in C(1:lastv,:). */
lastc = iladlc_(&lastv, n, &c__[c_offset], ldc);
} else {
/* Scan for the last non-zero row in C(:,1:lastv). */
lastc = iladlr_(m, &lastv, &c__[c_offset], ldc);
}

Thus when compiling qlarf.c the C compiler thinks iladlc_ takes a quadreal when in fact it takes a double and
errors occur at runtime in the call.

The fix would be to have the script generate the q and w versions of ila*lc.c ila*lr.c and when generating
the quad q*.c and w*.c files, replace the calls to iladlc_ iladlr_ (and z version) with ilqdlc_ ilqdlr_ (and
z versions)

show more ...

go back to the use of MPICH that I changed wrongly in 7904a33

fix %d that should be %D