Merge pull request #7441 from raffenet/4.3.x-fixes

[4.3.x] a batch of backported fixes

Approved-by: Hui Zhou <[email protected]>

Author: raffenet

CHANGES

@@ -1,3 +1,37 @@
+===============================================================================
+                               Changes in 4.3.1
+===============================================================================
+# Fix initialization in GPU-aware builds when no devices are present
+
+# Fix internal pmix.h header conflict when building with an external
+  PMIx library.
+
+# Fix build issue with Slurm by removing dependency on libslurm and
+  always using internal logic for parsing the Slurm hostfile.
+
+# Fix potential stale GPU IPC handle usage resulting in data corruption
+  or crashes
+
+# Update XPMEM thresholds to avoid excessive buffer mapping overhead
+
+# Fix potential hang in ROMIO when setting info hints on certain files
+
+# Improved detection of incompatible PMI[x] client/server configuration
+
+# Fix use of PMIX_PREFIX attribute for certain versions of OpenPMIx
+
+# Fix Intel GPU output with MPIR_CVAR_DEBUG_SUMMARY
+
+# Fix F08 binding compilation with nvfortran
+
+# Fix line continuation Hydra's --configfile option
+
+# Fix valgrind uninitialized read warnings in ch3
+
+# Fix missing mpixxx_opts.conf file with help text for mpicc and friends
+
+# Fixes for several compiler warnings
+
 ===============================================================================
                                Changes in 4.3.0
 ===============================================================================

Makefile.am

@@ -163,7 +163,7 @@ if BUILD_PROFILING_LIB
 # don't contribute any PMPI symbols.
 lib_LTLIBRARIES += lib/lib@[email protected]
 lib_lib@PMPIABILIBNAME@_la_SOURCES = $(mpi_abi_sources) $(mpi_f77_sources) $(mpi_core_sources)
-lib_lib@PMPIABILIBNAME@_la_LDFLAGS = $(external_ldflags) $(ABIVERSIONFLAGS)
+lib_lib@PMPIABILIBNAME@_la_LDFLAGS = $(external_ldflags) $(ABI_ABIVERSIONFLAGS)
 lib_lib@PMPIABILIBNAME@_la_CPPFLAGS = $(AM_CPPFLAGS) -DF77_USE_PMPI $(abi_cppflags)
 lib_lib@PMPIABILIBNAME@_la_LIBADD = $(external_libs) $(pmpi_convenience_libs)
 EXTRA_lib_lib@PMPIABILIBNAME@_la_DEPENDENCIES = $(pmpi_convenience_libs)
@@ -173,7 +173,7 @@ EXTRA_lib_lib@PMPIABILIBNAME@_la_DEPENDENCIES = $(pmpi_convenience_libs)
 # confused. (see https://bugzilla.redhat.com/show_bug.cgi?id=91110)
 lib_LTLIBRARIES += lib/lib@[email protected]
 lib_lib@MPIABILIBNAME@_la_SOURCES = $(mpi_abi_sources)
-lib_lib@MPIABILIBNAME@_la_LDFLAGS = $(ABIVERSIONFLAGS)
+lib_lib@MPIABILIBNAME@_la_LDFLAGS = $(ABI_ABIVERSIONFLAGS)
 lib_lib@MPIABILIBNAME@_la_CPPFLAGS = $(AM_CPPFLAGS) -DMPICH_MPI_FROM_PMPI $(abi_cppflags)
 lib_lib@MPIABILIBNAME@_la_LIBADD = lib/lib@[email protected] $(mpi_convenience_libs)
 
@@ -185,7 +185,7 @@ else !BUILD_PROFILING_LIB
 
 lib_LTLIBRARIES += lib/lib@[email protected]
 lib_lib@MPIABILIBNAME@_la_SOURCES = $(mpi_abi_sources) $(mpi_core_sources)
-lib_lib@MPIABILIBNAME@_la_LDFLAGS = $(external_ldflags) $(ABIVERSIONFLAGS)
+lib_lib@MPIABILIBNAME@_la_LDFLAGS = $(external_ldflags) $(ABI_ABIVERSIONFLAGS)
 lib_lib@MPIABILIBNAME@_la_CPPFLAGS = $(AM_CPPFLAGS) $(abi_cppflags)
 lib_lib@MPIABILIBNAME@_la_LIBADD = $(external_libs) $(pmpi_convenience_libs) $(mpi_convenience_libs)
 EXTRA_lib_lib@MPIABILIBNAME@_la_DEPENDENCIES = $(pmpi_convenience_libs) $(mpi_convenience_libs)

confdb/aclocal_fc.m4

@@ -1268,6 +1268,44 @@ AC_DEFUN([PAC_FC_CHECK_IGNORE_TKR],[
     AC_LANG_POP(Fortran)
 ])
 
+dnl
+dnl PAC_FC_CHECK_IGNORE_TKR_D check directives to ignore "DEVICE" attribute.
+dnl set pac_fc_ignore_tkr_d to a type if supported, otherwise, no.
+dnl
+AC_DEFUN([PAC_FC_CHECK_IGNORE_TKR_D],[
+    AC_LANG_PUSH(Fortran)
+    AC_MSG_CHECKING([directives for Fortran compiler to ignore TKR(d) check])
+    pac_fc_ignore_tkr_d=no
+    for a in dir ; do
+        case $a in
+            dir)
+                # flang
+                decl='!DIR$ IGNORE_TKR(d) buf'
+                ;;
+        esac
+
+        AC_COMPILE_IFELSE([AC_LANG_SOURCE([
+            program main
+                IMPLICIT NONE
+                INTERFACE
+                  SUBROUTINE FUNC_A(buf)
+                    TYPE(*), DIMENSION(..) :: buf
+                    $decl
+                  END SUBROUTINE
+                END INTERFACE
+
+                REAL A(10)
+                CALL FUNC_A(A)
+            end
+        ])],[pac_fc_ignore_tkr_d=$a],[])
+        if test $pac_fc_ignore_tkr_d != no ; then
+            break
+        fi
+    done
+    AC_MSG_RESULT([$pac_fc_ignore_tkr_d])
+    AC_LANG_POP(Fortran)
+])
+
 dnl
 dnl PAC_FC_ISO_C_BINDING check whether ISO_C_BINDING is supported.
 dnl set pac_fc_iso_c_binding to yes if it's supported, otherwise, no.

configure.ac

@@ -495,8 +495,10 @@ AC_ARG_ENABLE(versioning,
 
 if test "$enable_versioning" = "yes" ; then
    ABIVERSIONFLAGS="-version-info \$(ABIVERSION)"
+   ABI_ABIVERSIONFLAGS="-version-info libmpi_abi_so_verion_m4"
 else
    ABIVERSIONFLAGS="-avoid-version"
+   ABI_ABIVERSIONFLAGS="-avoid-version"
 fi
 export ABIVERSIONFLAGS
 AC_SUBST(ABIVERSIONFLAGS)
@@ -4212,6 +4214,7 @@ fi
 
 if test "$enable_f08" = "yes" ; then
     PAC_FC_CHECK_REAL128
+    PAC_FC_CHECK_IGNORE_TKR_D
     if test -z "$PYTHON" ; then
         if test -f src/binding/fortran/use_mpi_f08/mpi_f08.f90 ; then
             AC_MSG_NOTICE([Using pre-generated Fortran mpi_f08 binding source. To prevent issues, install Python 3 and rerun configure.])
@@ -4223,6 +4226,9 @@ if test "$enable_f08" = "yes" ; then
         if test "$pac_fc_has_real128" = "no" ; then
             cmd_f08="$cmd_f08 -no-real128"
         fi
+        if test "$pac_fc_ignore_tkr_d" != "no" ; then
+            cmd_f08="$cmd_f08 -ignore-tkr-d=$pac_fc_ignore_tkr_d"
+        fi
         if test "$enable_romio" = "no" ; then
             cmd_f08="$cmd_f08 -no-mpiio"
         fi

doc/wiki/source_code/Creating_A_Release.md

@@ -178,7 +178,7 @@ cp -v path/to/{mpich,hydra,libpmi,mpich-testsuite}-X.Y.Z.tar.gz downloads/X.Y.Z/
     Shortlog can be generated with, e.g.,
 
 ```
-git log --no-merges --format="format:[%cd] %s" --date=short v3.4.2..v4.0a2 > shortlog
+git log --no-merges --format=reference --date=short v4.2.3..v4.3.0 > shortlog
 ```
 
   - For full release update documentation as well: README.txt,

maint/extracterrmsgs

@@ -875,7 +875,7 @@ sub ExpandDir {
 	elsif (-d "$dir/$filename") {
 	    $otherdirs[$#otherdirs+1] = "$dir/$filename";
 	}
-	elsif ($filename =~ /(.*\.[chi])(pp){0,1}$/) {
+	elsif ($filename =~ /(.*\.([chi])(pp){0,1}|inc)$/) {
 	    # Test for both Unix- and Windows-style directory separators
 	    if (!defined($skipFiles{"$dir/$filename"}) &&
 		!defined($skipFiles{"$dir\\$filename"})) {

maint/local_python/binding_c.py

@@ -1106,7 +1106,7 @@ def out_can_be_undefined(p):
         G.out.append("int ret = " + static_call + ";")
         for l in post_filters:
             G.out.append(l)
-        if re.match(r'MPI_(Init|Init_thread|Session_init)$', func_name, re.IGNORECASE):
+        if re.match(r'MPI_(Init|Init_thread|Session_init)|MPI_T_init_thread$', func_name, re.IGNORECASE):
             G.out.append("ABI_init_builtins();")
         G.out.append("return ret;")
     G.out.append("DEDENT")

maint/local_python/binding_f08.py

@@ -879,6 +879,10 @@ def dump_interface_function(func, name, c_name, is_large):
 def dump_mpi_f08(func, is_large):
     f08_mapping = get_kind_map('F08', is_large)
 
+    def dump_ignore_tkr_d(name, out):
+        if G.opts['ignore-tkr-d'] == 'dir': # e.g. nvfortran
+            out.append("!DIR$ IGNORE_TKR(d) " + name)
+
     uses = {}
     f_param_list = []
     decl_list = []
@@ -889,6 +893,8 @@ def dump_mpi_f08(func, is_large):
         f_param_list.append(p['name'])
         decl = get_F_decl(p, f08_mapping)
         decl_list.append(decl)
+        if 'ignore-tkr-d' in G.opts and p['kind'] == 'BUFFER':
+            dump_ignore_tkr_d(p['name'], decl_list)
         check_decl_uses(decl, uses)
     if 'return' not in func:
         f_param_list.append("ierror")

maint/local_python/info_hints.py

@@ -10,7 +10,12 @@ def collect_info_hint_blocks(root_dir):
     """Collect INFO_HINT_BLOCKS from source files and return a function-name-keyed dictionary"""
     infos_by_funcname = {}
 
-    files = subprocess.check_output("find %s -name '*.[ch]' |xargs grep -l BEGIN_INFO_HINT_BLOCK" % root_dir, shell=True).splitlines()
+    # xargs may process files in batches and return a non-zero exit code if grep comes up empty
+    # use "|| true" to avoid terminating the script in those instances
+    # https://stackoverflow.com/questions/26540813/got-exit-code-123-in-find-xargs-grep
+    files = subprocess.check_output(
+        "find %s -name '*.[ch]' |xargs grep -l BEGIN_INFO_HINT_BLOCK || true" % root_dir,
+        shell=True).splitlines()
     for f in files:
         infos = parse_info_block(f)
         for info in infos:

maint/version.m4

@@ -14,7 +14,7 @@
 # changing this by playing with diversions, but then we would probably be
 # playing with autotools-fire.
 
-m4_define([MPICH_VERSION_m4],[4.3.0])dnl
+m4_define([MPICH_VERSION_m4],[4.3.1rc1])dnl
 m4_define([MPICH_RELEASE_DATE_m4],[unreleased development copy])dnl
 
 # For libtool ABI versioning rules see:
@@ -40,4 +40,7 @@ m4_define([MPICH_RELEASE_DATE_m4],[unreleased development copy])dnl
 # last version 4.3.0 - 17:0:5
 m4_define([libmpi_so_version_m4],[0:0:0])dnl
 
+# libmpi_abi.so version
+m4_define([libmpi_abi_so_version_m4],[1:0:0])dnl
+
 [#] end of __file__