Package: openmpi
Version: 1.3.3-2
Severity: important
Tags: patch
User: [email protected]
Usertags: hurd
Hi,
currently openmpi does not build on GNU/Hurd for various small reasons, like
unconditional usage of PATH_MAX, MAXPATHLEN, and MAXHOSTNAMELEN, usage of
sigaction::sa_sigaction, and a wrong header check for recognizing Darwin.
The attached patch hopefully fixes all the issues:
- missing SA_SIGINFO, so a sa_handler is used; show_stackframe_handler() calls
show_stackframe() with NULL parameters as that handler checks for non-NULL
siginfo_t* and context* already
- some fixed-size char[] for getcwd() to get_current_dir_name() + free()
- some fixed-size char[] for gethostname() to a dynamic realloc() loop
- one gethostname() call commented out, as it seems unused...
- look for mach-o/arch.h instead of mach/mach_host.h to determine whether it is
compiling on Darwin
- one malloc() + free() instead of a fixed-size char[PATH_MAX]
Please note that, given a .m4 macro is changed, then an autoreconf'ing of the
build system is needed.
Thanks,
--
Pino
--- openmpi-1.3.3.orig/opal/util/stacktrace.c
+++ openmpi-1.3.3/opal/util/stacktrace.c
@@ -363,6 +363,13 @@
fflush(stderr);
}
+#ifndef SA_SIGINFO
+static void show_stackframe_handler (int signo)
+{
+ show_stackframe(signo, NULL, NULL);
+}
+
+#endif /* SA_SIGINFO */
#endif /* OMPI_WANT_PRETTY_PRINT_STACKTRACE && ! defined(__WINDOWS__) */
@@ -422,8 +429,12 @@
mca_base_param_lookup_string (param, &string_value);
memset(&act, 0, sizeof(act));
+#ifdef SA_SIGINFO
act.sa_sigaction = show_stackframe;
act.sa_flags = SA_SIGINFO;
+#else
+ act.sa_handler = show_stackframe_handler;
+#endif
#ifdef SA_ONESHOT
act.sa_flags |= SA_ONESHOT;
#else
--- openmpi-1.3.3.orig/opal/mca/base/mca_base_param.c
+++ openmpi-1.3.3/opal/mca/base/mca_base_param.c
@@ -186,8 +186,14 @@
home = (char*)opal_home_directory();
if(NULL == cwd) {
+#if !defined(MAXPATHLEN) && defined(__GLIBC__)
+ cwd = get_current_dir_name();
+ if( NULL == cwd)
+#else
cwd = (char *) malloc(sizeof(char) * MAXPATHLEN);
- if( NULL == (cwd = getcwd(cwd, MAXPATHLEN) )) {
+ if( NULL == (cwd = getcwd(cwd, MAXPATHLEN) ))
+#endif
+ {
opal_output(0, "Error: Unable to get the current working directory\n");
cwd = strdup(".");
}
--- openmpi-1.3.3.orig/orte/mca/odls/base/odls_base_default_fns.c
+++ openmpi-1.3.3/orte/mca/odls/base/odls_base_default_fns.c
@@ -750,8 +750,13 @@
orte_odls_job_t *jobdat;
orte_pmap_t *pmap;
char *pathenv = NULL, *mpiexec_pathenv = NULL;
+#if !defined(MAXPATHLEN) && defined(__GLIBC__)
+ char *basedir=NULL;
+ char *dir=NULL;
+#else
char basedir[MAXPATHLEN];
char dir[MAXPATHLEN];
+#endif
char **argvptr;
char *full_search;
char **argvsav=NULL;
@@ -764,7 +769,11 @@
* bouncing around as we execute various apps, but we will always return
* to this place as our default directory
*/
+#if !defined(MAXPATHLEN) && defined(__GLIBC__)
+ basedir = get_current_dir_name();
+#else
getcwd(basedir, sizeof(basedir));
+#endif
/* find the jobdat for this job */
jobdat = NULL;
@@ -915,7 +924,11 @@
* again not match getcwd! This is beyond our control - we are only
* ensuring they start out matching.
*/
+#if !defined(MAXPATHLEN) && defined(__GLIBC__)
+ dir = get_current_dir_name();
+#else
getcwd(dir, sizeof(dir));
+#endif
opal_setenv("PWD", dir, true, &app->env);
/* Search for the OMPI_exec_path and PATH settings in the environment. */
@@ -1247,6 +1260,10 @@
opal_condition_signal(&orte_odls_globals.cond);
OPAL_THREAD_UNLOCK(&orte_odls_globals.mutex);
+#if !defined(MAXPATHLEN) && defined(__GLIBC__)
+ free(basedir);
+ free(dir);
+#endif
return rc;
}
--- openmpi-1.3.3.orig/opal/mca/paffinity/darwin/configure.m4
+++ openmpi-1.3.3/opal/mca/paffinity/darwin/configure.m4
@@ -23,9 +23,9 @@
# -----------------------------------------------------------
AC_DEFUN([MCA_paffinity_darwin_CONFIG],[
OMPI_VAR_SCOPE_PUSH([paff_darwin_happy])
- # check to see if we have <mach/mach_host.h>
+ # check to see if we have <mach-o/arch.h>
# as this is a Darwin-specific thing
- AC_CHECK_HEADER([mach/mach_host.h], [paff_darwin_happy=yes], [paff_darwin_happy=no])
+ AC_CHECK_HEADER([mach-o/arch.h], [paff_darwin_happy=yes], [paff_darwin_happy=no])
AS_IF([test "$paff_darwin_happy" = "yes"], [$1], [$2])
OMPI_VAR_SCOPE_POP
--- openmpi-1.3.3.orig/opal/mca/base/mca_base_component_find.c
+++ openmpi-1.3.3/opal/mca/base/mca_base_component_find.c
@@ -201,11 +201,16 @@
}
if (opal_list_get_end(found_components) == item) {
- char h[MAXHOSTNAMELEN];
- gethostname(h, sizeof(h));
+ char *h = NULL;
+ size_t h_length = 128;
+ do {
+ h_length *= 2;
+ h = realloc(h, h_length);
+ } while ((gethostname(h, h_length) == -1) && (errno == ENAMETOOLONG));
opal_show_help("help-mca-base.txt",
"find-available:not-valid", true,
h, type, requested_component_names[i]);
+ free(h);
return OPAL_ERR_NOT_FOUND;
}
}
--- openmpi-1.3.3.orig/orte/util/context_fns.c
+++ openmpi-1.3.3/orte/util/context_fns.c
@@ -56,10 +56,12 @@
{
bool good = true;
const char *tmp;
+#if 0 /* 'hostname' looks unused... */
char hostname[MAXHOSTNAMELEN];
/* Use hostname in a few messages below */
gethostname(hostname, sizeof(hostname));
+#endif
/* If we want to chdir and the chdir fails (for any reason -- such
as if the dir doesn't exist, it isn't a dir, we don't have
--- openmpi-1.3.3.orig/ompi/runtime/ompi_mpi_finalize.c
+++ openmpi-1.3.3/ompi/runtime/ompi_mpi_finalize.c
@@ -100,13 +100,18 @@
/* Note that if we're already finalized, we cannot raise an
MPI exception. The best that we can do is write something
to stderr. */
- char hostname[MAXHOSTNAMELEN];
+ char *hostname = NULL;
+ size_t hostname_length = 128;
pid_t pid = getpid();
- gethostname(hostname, sizeof(hostname));
+ do {
+ hostname_length *= 2;
+ hostname = realloc(hostname, hostname_length);
+ } while ((gethostname(hostname, hostname_length) == -1) && (errno == ENAMETOOLONG));
orte_show_help("help-mpi-runtime.txt",
"mpi_finalize:invoked_multiple_times",
true, hostname, pid);
+ free(hostname);
return MPI_ERR_OTHER;
}
--- openmpi-1.3.3.orig/ompi/runtime/ompi_mpi_abort.c
+++ openmpi-1.3.3/ompi/runtime/ompi_mpi_abort.c
@@ -53,10 +53,11 @@
bool kill_remote_of_intercomm)
{
int count = 0, i;
- char *msg, *host, hostname[MAXHOSTNAMELEN];
+ char *msg, *host = NULL;
pid_t pid = 0;
orte_process_name_t *abort_procs;
orte_std_cntr_t nabort_procs;
+ bool free_host = false;
/* Protection for recursive invocation */
if (have_been_invoked) {
@@ -70,8 +71,12 @@
if (orte_initialized) {
host = orte_process_info.nodename;
} else {
- gethostname(hostname, sizeof(hostname));
- host = hostname;
+ size_t host_length = 128;
+ do {
+ host_length *= 2;
+ host = realloc(host, host_length);
+ } while ((gethostname(host, host_length) == -1) && (errno == ENAMETOOLONG));
+ free_host = true;
}
pid = getpid();
@@ -138,9 +143,17 @@
fprintf(stderr, "[%s:%d] Abort %s completed successfully; not able to guarantee that all other processes were killed!\n",
host, (int) pid, ompi_mpi_finalized ?
"after MPI_FINALIZE" : "before MPI_INIT");
+ if (free_host) {
+ free(host);
+ }
exit(errcode);
}
+ if (free_host) {
+ free(host);
+ host = NULL;
+ }
+
/* abort local procs in the communicator. If the communicator is
an intercommunicator AND the abort has explicitly requested
that we abort the remote procs, then do that as well. */
--- openmpi-1.3.3.orig/orte/orted/orted_main.c
+++ openmpi-1.3.3/orte/orted/orted_main.c
@@ -197,7 +197,7 @@
int ret = 0;
int fd;
opal_cmd_line_t *cmd_line = NULL;
- char log_file[PATH_MAX];
+ char *log_file = NULL;
char *jobidstring;
char *rml_uri;
int i;
@@ -439,6 +439,7 @@
}
/* define a log file name in the session directory */
+ log_file = malloc(strlen(jobidstring) + strlen(orte_process_info.nodename) + 19);
sprintf(log_file, "output-orted-%s-%s.log",
jobidstring, orte_process_info.nodename);
log_path = opal_os_path(false,
@@ -446,6 +447,7 @@
orte_process_info.top_session_dir,
log_file,
NULL);
+ free(log_file);
fd = open(log_path, O_RDWR|O_CREAT|O_TRUNC, 0640);
if (fd < 0) {
