[svn-r4411] Fixed nemesis to correctly set vc state (instead of setting all vcs to active.  Renamed MPIDI_Comm_get_vc to MPIDI_Comm_get_vc_set_active to alert caller to side-effect.  Added MPIDI_CHANGE_VC_STATE macro to set the vc state and call debugging macro.  Changed all places where vc state is changed to use this macro.  Reviewed by goodell@.

[svn-r4410] Adding rules for building fortran 90 modules to windows makefile/scripts & Adding the module files to the 64-bit installer

correct and ensures none of the fs-specific open routines will hang if they try
to do anything clever (like broadcast stat info to all processors)

processes with a single stat system call

Added AC_CONFIG_AUX_DIR in rlog's configure.in to get rid of install-sh.

[svn-r4405] Adding the windows header file for NT intrinsics - to define the non-intrinsic windows data types. Without this fix the makefile builds fail after r4322.

downstream connections in the future.

calls.

[svn-r4402] Correcting the format of integer model & integer model map in windows configure. The integer model is *not* a single integer but a series in "r1, k1, r2, k2,..." format & both integer model and integer model map end with a comma separator.

[svn-r4401] fixed the rlog configure.in and Makefile.sm as well as top-level configure.in so rlog tools can be installed in $bindir.  Fixed for req#497.

-Thislline, and those below, will be ignored--

M    configure.in
M    src/util/logging/rlog/configure.in
M    src/util/logging/rlog/Makefile.sm
AM   src/util/logging/install-sh

[svn-r4398] Adding F90 type create funcs - part of Extended fortran support - to fortran libs on windows. The 32-bit fortran libs should have the complete extented fortran support now.

temporarily break the PMI-2 client code, but is the general direction
we want to change the client code in.

build to fail with --enable-g. This should fix ticket #569.

you printf a variable without saying it's a string.

reviewd-by: goodell

[svn-r4388] In the case where the signal raised by mpiexec does not terminate it, we were having an infinite loop.

trying to convert it to the correct format (different UI
implementations might want to do different things with the actual
return code).

2. Modified the mpiexec UI to separate out return() or exit() kind of
return codes and signal-based terminations. In the case of
signal-based terminations, mpiexec just raises the same signal and
aborts. This allows the shell to find the correct string to print on
the screen. Plus, this allows debuggers to catch the signal and do a
trace dump as well.

This should fix ticket #518.

.  Since it is an abort instead of warning, made the error message more explicit of what compilers and compiler flags are being used.

[svn-r4383] Added --enable-check-compiler-flags option to allow user to disable the checks of user-supplied compiler flasg, xxFLAGS and MPICH2LIB_xxFLAGS.  This may be necessary as compiler like Portland Group, PG-7 or PG-8, does not like PAC_*_CHECK_COMPILER_OPTION.

directory).

[svn-r4377] removed dead and invalid PMI routines from pmi.h and from the simple and smpd PMI implementations; also added necessary include file for smpd to compile under Mac OSX

[svn-r4375] Added note about the unexpected reqreuiement to use -cpp with the IBM mpCC compiler to actually get it to use the MPI C++ bindings (!!!)