1
1
openmpi/opal/util/output.c

980 строки
26 KiB
C
Исходник Обычный вид История

/*
* Copyright (c) 2004-2010 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2008 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2006 High Performance Computing Center Stuttgart,
* University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2006 The Regents of the University of California.
* All rights reserved.
* Copyright (c) 2007-2008 Cisco Systems, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#include "opal_config.h"
#include <stdio.h>
#include <stdarg.h>
#include <stdlib.h>
#ifdef HAVE_SYSLOG_H
#include <syslog.h>
#endif
#include <string.h>
#include <fcntl.h>
#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif
#ifdef HAVE_SYS_PARAM_H
#include <sys/param.h>
#endif
#include "opal/util/opal_environ.h"
#include "opal/util/output.h"
#include "opal/threads/mutex.h"
#include "opal/constants.h"
/*
* Private data
*/
static int verbose_stream = -1;
static opal_output_stream_t verbose;
static char *output_dir = NULL;
static char *output_prefix = NULL;
/*
* Internal data structures and helpers for the generalized output
* stream mechanism.
*/
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
typedef struct {
bool ldi_used;
bool ldi_enabled;
int ldi_verbose_level;
bool ldi_syslog;
int ldi_syslog_priority;
char *ldi_syslog_ident;
char *ldi_prefix;
int ldi_prefix_len;
char *ldi_suffix;
int ldi_suffix_len;
bool ldi_stdout;
bool ldi_stderr;
bool ldi_file;
bool ldi_file_want_append;
char *ldi_file_suffix;
int ldi_fd;
int ldi_file_num_lines_lost;
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
} output_desc_t;
/*
* Private functions
*/
static void construct(opal_object_t *stream);
static int do_open(int output_id, opal_output_stream_t * lds);
static int open_file(int i);
static void free_descriptor(int output_id);
static int make_string(char **no_newline_string, output_desc_t *ldi,
const char *format, va_list arglist);
static int output(int output_id, const char *format, va_list arglist);
#define OPAL_OUTPUT_MAX_STREAMS 64
#if defined(HAVE_SYSLOG)
#define USE_SYSLOG 1
#else
#define USE_SYSLOG 0
#endif
/* global state */
bool opal_output_redirected_to_syslog = false;
int opal_output_redirected_syslog_pri;
/*
* Local state
*/
static bool initialized = false;
static int default_stderr_fd = -1;
static output_desc_t info[OPAL_OUTPUT_MAX_STREAMS];
static char *temp_str = 0;
static size_t temp_str_len = 0;
static opal_mutex_t mutex;
#if defined(HAVE_SYSLOG)
static bool syslog_opened = false;
#endif
static char *redirect_syslog_ident = NULL;
OBJ_CLASS_INSTANCE(opal_output_stream_t, opal_object_t, construct, NULL);
/*
* Setup the output stream infrastructure
*/
bool opal_output_init(void)
{
int i;
char hostname[32];
char *str;
if (initialized) {
return true;
}
str = getenv("OPAL_OUTPUT_STDERR_FD");
if (NULL != str) {
default_stderr_fd = atoi(str);
}
str = getenv("OPAL_OUTPUT_REDIRECT");
if (NULL != str) {
if (0 == strcasecmp(str, "syslog")) {
opal_output_redirected_to_syslog = true;
}
}
str = getenv("OPAL_OUTPUT_SYSLOG_PRI");
if (NULL != str) {
if (0 == strcasecmp(str, "info")) {
opal_output_redirected_syslog_pri = LOG_INFO;
} else if (0 == strcasecmp(str, "error")) {
opal_output_redirected_syslog_pri = LOG_ERR;
} else if (0 == strcasecmp(str, "warn")) {
opal_output_redirected_syslog_pri = LOG_WARNING;
} else {
opal_output_redirected_syslog_pri = LOG_ERR;
}
} else {
opal_output_redirected_syslog_pri = LOG_ERR;
}
str = getenv("OPAL_OUTPUT_SYSLOG_IDENT");
if (NULL != str) {
redirect_syslog_ident = strdup(str);
}
OBJ_CONSTRUCT(&verbose, opal_output_stream_t);
if (opal_output_redirected_to_syslog) {
verbose.lds_want_syslog = true;
verbose.lds_syslog_priority = opal_output_redirected_syslog_pri;
if (NULL != str) {
verbose.lds_syslog_ident = strdup(redirect_syslog_ident);
}
verbose.lds_want_stderr = false;
verbose.lds_want_stdout = false;
} else {
verbose.lds_want_stderr = true;
}
gethostname(hostname, sizeof(hostname));
hostname[sizeof(hostname)-1] = '\0';
asprintf(&verbose.lds_prefix, "[%s:%05d] ", hostname, getpid());
for (i = 0; i < OPAL_OUTPUT_MAX_STREAMS; ++i) {
info[i].ldi_used = false;
info[i].ldi_enabled = false;
info[i].ldi_syslog = opal_output_redirected_to_syslog;
info[i].ldi_file = false;
info[i].ldi_file_suffix = NULL;
info[i].ldi_file_want_append = false;
info[i].ldi_fd = -1;
info[i].ldi_file_num_lines_lost = 0;
}
/* Initialize the mutex that protects the output */
OBJ_CONSTRUCT(&mutex, opal_mutex_t);
initialized = true;
/* Set some defaults */
asprintf(&output_prefix, "output-pid%d-", getpid());
output_dir = strdup(opal_tmp_directory());
/* Open the default verbose stream */
verbose_stream = opal_output_open(&verbose);
return true;
}
/*
* Open a stream
*/
int opal_output_open(opal_output_stream_t * lds)
{
return do_open(-1, lds);
}
/*
* Reset the parameters on a stream
*/
int opal_output_reopen(int output_id, opal_output_stream_t * lds)
{
return do_open(output_id, lds);
}
/*
* Enable and disable output streams
*/
bool opal_output_switch(int output_id, bool enable)
{
bool ret = false;
/* Setup */
if (!initialized) {
opal_output_init();
}
if (output_id >= 0 && output_id < OPAL_OUTPUT_MAX_STREAMS) {
ret = info[output_id].ldi_enabled;
info[output_id].ldi_enabled = enable;
}
return ret;
}
/*
* Reopen all the streams; used during checkpoint/restart.
*/
void opal_output_reopen_all(void)
{
char *str;
char hostname[32];
str = getenv("OPAL_OUTPUT_STDERR_FD");
if (NULL != str) {
default_stderr_fd = atoi(str);
} else {
default_stderr_fd = -1;
}
gethostname(hostname, sizeof(hostname));
if( NULL != verbose.lds_prefix ) {
free(verbose.lds_prefix);
verbose.lds_prefix = NULL;
}
asprintf(&verbose.lds_prefix, "[%s:%05d] ", hostname, getpid());
#if 0
int i;
opal_output_stream_t lds;
for (i = 0; i < OPAL_OUTPUT_MAX_STREAMS; ++i) {
/* scan till we find ldi_used == 0, which is the end-marker */
if (!info[i].ldi_used) {
break;
}
/*
* set this to zero to ensure that opal_output_open will
* return this same index as the output stream id
*/
info[i].ldi_used = false;
#if USE_SYSLOG
lds.lds_want_syslog = info[i].ldi_syslog;
lds.lds_syslog_priority = info[i].ldi_syslog_priority;
lds.lds_syslog_ident = info[i].ldi_syslog_ident;
#else
lds.lds_want_syslog = false;
#endif
lds.lds_prefix = info[i].ldi_prefix;
lds.lds_suffix = info[i].ldi_suffix;
lds.lds_want_stdout = info[i].ldi_stdout;
lds.lds_want_stderr = info[i].ldi_stderr;
lds.lds_want_file = (-1 == info[i].ldi_fd) ? false : true;
/* open all streams in append mode */
lds.lds_want_file_append = true;
lds.lds_file_suffix = info[i].ldi_file_suffix;
/*
* call opal_output_open to open the stream. The return value
* is guaranteed to be i. So we can ignore it.
*/
opal_output_open(&lds);
}
#endif
}
/*
* Close a stream
*/
void opal_output_close(int output_id)
{
int i;
/* Setup */
if (!initialized) {
return;
}
/* If it's valid, used, enabled, and has an open file descriptor,
* free the resources associated with the descriptor */
OPAL_THREAD_LOCK(&mutex);
if (output_id >= 0 && output_id < OPAL_OUTPUT_MAX_STREAMS &&
info[output_id].ldi_used && info[output_id].ldi_enabled) {
free_descriptor(output_id);
/* If no one has the syslog open, we should close it */
for (i = 0; i < OPAL_OUTPUT_MAX_STREAMS; ++i) {
if (info[i].ldi_used && info[i].ldi_syslog) {
break;
}
}
#if defined(HAVE_SYSLOG)
if (i >= OPAL_OUTPUT_MAX_STREAMS && syslog_opened) {
closelog();
}
#endif
}
OPAL_THREAD_UNLOCK(&mutex);
}
/*
* Main function to send output to a stream
*/
void opal_output(int output_id, const char *format, ...)
{
if (output_id >= 0 && output_id < OPAL_OUTPUT_MAX_STREAMS) {
va_list arglist;
va_start(arglist, format);
output(output_id, format, arglist);
va_end(arglist);
}
}
/*
* Send a message to a stream if the verbose level is high enough
*/
void opal_output_verbose(int level, int output_id, const char *format, ...)
{
if (output_id >= 0 && output_id < OPAL_OUTPUT_MAX_STREAMS &&
info[output_id].ldi_verbose_level >= level) {
va_list arglist;
va_start(arglist, format);
output(output_id, format, arglist);
va_end(arglist);
}
}
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
/*
* Send a message to a stream if the verbose level is high enough
*/
void opal_output_vverbose(int level, int output_id, const char *format,
va_list arglist)
{
if (output_id >= 0 && output_id < OPAL_OUTPUT_MAX_STREAMS &&
info[output_id].ldi_verbose_level >= level) {
output(output_id, format, arglist);
}
}
/*
* Send a message to a string if the verbose level is high enough
*/
char *opal_output_string(int level, int output_id, const char *format, ...)
{
int rc;
char *ret = NULL;
if (output_id >= 0 && output_id < OPAL_OUTPUT_MAX_STREAMS &&
info[output_id].ldi_verbose_level >= level) {
va_list arglist;
va_start(arglist, format);
rc = make_string(&ret, &info[output_id], format, arglist);
va_end(arglist);
if (OPAL_SUCCESS != rc) {
ret = NULL;
}
}
return ret;
}
/*
* Send a message to a string if the verbose level is high enough
*/
char *opal_output_vstring(int level, int output_id, const char *format,
va_list arglist)
{
int rc;
char *ret = NULL;
if (output_id >= 0 && output_id < OPAL_OUTPUT_MAX_STREAMS &&
info[output_id].ldi_verbose_level >= level) {
rc = make_string(&ret, &info[output_id], format, arglist);
if (OPAL_SUCCESS != rc) {
ret = NULL;
}
}
return ret;
}
/*
* Set the verbosity level of a stream
*/
void opal_output_set_verbosity(int output_id, int level)
{
if (output_id >= 0 && output_id < OPAL_OUTPUT_MAX_STREAMS) {
info[output_id].ldi_verbose_level = level;
}
}
/*
* Control where output flies will go
*/
void opal_output_set_output_file_info(const char *dir,
const char *prefix,
char **olddir,
char **oldprefix)
{
if (NULL != olddir) {
*olddir = strdup(output_dir);
}
if (NULL != oldprefix) {
*oldprefix = strdup(output_prefix);
}
if (NULL != dir) {
free(output_dir);
output_dir = strdup(dir);
}
if (NULL != prefix) {
free(output_prefix);
output_prefix = strdup(prefix);
}
}
/*
* Shut down the output stream system
*/
void opal_output_finalize(void)
{
if (initialized) {
if (verbose_stream != -1) {
opal_output_close(verbose_stream);
}
free(verbose.lds_prefix);
verbose_stream = -1;
free (output_prefix);
free (output_dir);
if(NULL != temp_str) {
free(temp_str);
temp_str = NULL;
temp_str_len = 0;
}
OBJ_DESTRUCT(&verbose);
OBJ_DESTRUCT(&mutex);
}
}
/************************************************************************/
/*
* Constructor
*/
static void construct(opal_object_t *obj)
{
opal_output_stream_t *stream = (opal_output_stream_t*) obj;
stream->lds_verbose_level = 0;
stream->lds_syslog_priority = 0;
stream->lds_syslog_ident = NULL;
stream->lds_prefix = NULL;
stream->lds_suffix = NULL;
stream->lds_is_debugging = false;
stream->lds_want_syslog = false;
stream->lds_want_stdout = false;
stream->lds_want_stderr = false;
stream->lds_want_file = false;
stream->lds_want_file_append = false;
stream->lds_file_suffix = NULL;
}
/*
* Back-end of open() and reopen(). Necessary to have it as a
* back-end function so that we can do the thread locking properly
* (especially upon reopen).
*/
static int do_open(int output_id, opal_output_stream_t * lds)
{
int i;
bool redirect_to_file = false;
char *str, *sfx;
/* Setup */
if (!initialized) {
opal_output_init();
}
str = getenv("OPAL_OUTPUT_REDIRECT");
if (NULL != str && 0 == strcasecmp(str, "file")) {
redirect_to_file = true;
}
sfx = getenv("OPAL_OUTPUT_SUFFIX");
/* If output_id == -1, find an available stream, or return
* OPAL_ERROR */
if (-1 == output_id) {
OPAL_THREAD_LOCK(&mutex);
for (i = 0; i < OPAL_OUTPUT_MAX_STREAMS; ++i) {
if (!info[i].ldi_used) {
break;
}
}
if (i >= OPAL_OUTPUT_MAX_STREAMS) {
OPAL_THREAD_UNLOCK(&mutex);
return OPAL_ERR_OUT_OF_RESOURCE;
}
}
/* Otherwise, we're reopening, so we need to free all previous
* resources, close files, etc. */
else {
free_descriptor(output_id);
i = output_id;
}
/* Special case: if we got NULL for lds, then just use the default
* verbose */
if (NULL == lds) {
lds = &verbose;
}
/* Got a stream -- now initialize it and open relevant outputs */
info[i].ldi_used = true;
if (-1 == output_id) {
OPAL_THREAD_UNLOCK(&mutex);
}
info[i].ldi_enabled = lds->lds_is_debugging ?
(bool) OPAL_ENABLE_DEBUG : true;
info[i].ldi_verbose_level = lds->lds_verbose_level;
#if USE_SYSLOG
#if defined(HAVE_SYSLOG)
if (opal_output_redirected_to_syslog) {
info[i].ldi_syslog = true;
info[i].ldi_syslog_priority = opal_output_redirected_syslog_pri;
if (NULL != redirect_syslog_ident) {
info[i].ldi_syslog_ident = strdup(redirect_syslog_ident);
openlog(redirect_syslog_ident, LOG_PID, LOG_USER);
} else {
info[i].ldi_syslog_ident = NULL;
openlog("opal", LOG_PID, LOG_USER);
}
syslog_opened = true;
} else {
#endif
info[i].ldi_syslog = lds->lds_want_syslog;
if (lds->lds_want_syslog) {
#if defined(HAVE_SYSLOG)
if (NULL != lds->lds_syslog_ident) {
info[i].ldi_syslog_ident = strdup(lds->lds_syslog_ident);
openlog(lds->lds_syslog_ident, LOG_PID, LOG_USER);
} else {
info[i].ldi_syslog_ident = NULL;
openlog("opal", LOG_PID, LOG_USER);
}
#endif
syslog_opened = true;
info[i].ldi_syslog_priority = lds->lds_syslog_priority;
}
#if defined(HAVE_SYSLOG)
}
#endif
#else
info[i].ldi_syslog = false;
#endif
if (NULL != lds->lds_prefix) {
info[i].ldi_prefix = strdup(lds->lds_prefix);
info[i].ldi_prefix_len = (int)strlen(lds->lds_prefix);
} else {
info[i].ldi_prefix = NULL;
info[i].ldi_prefix_len = 0;
}
if (NULL != lds->lds_suffix) {
info[i].ldi_suffix = strdup(lds->lds_suffix);
info[i].ldi_suffix_len = (int)strlen(lds->lds_suffix);
} else {
info[i].ldi_suffix = NULL;
info[i].ldi_suffix_len = 0;
}
if (opal_output_redirected_to_syslog) {
/* since all is redirected to syslog, ensure
* we don't duplicate the output to the std places
*/
info[i].ldi_stdout = false;
info[i].ldi_stderr = false;
info[i].ldi_file = false;
info[i].ldi_fd = -1;
} else {
/* since we aren't redirecting to syslog, use what was
* given to us
*/
if (NULL != str && redirect_to_file) {
info[i].ldi_stdout = false;
info[i].ldi_stderr = false;
info[i].ldi_file = true;
} else {
info[i].ldi_stdout = lds->lds_want_stdout;
info[i].ldi_stderr = lds->lds_want_stderr;
info[i].ldi_fd = -1;
info[i].ldi_file = lds->lds_want_file;
}
if (NULL != sfx) {
info[i].ldi_file_suffix = strdup(sfx);
} else {
info[i].ldi_file_suffix = (NULL == lds->lds_file_suffix) ? NULL :
strdup(lds->lds_file_suffix);
}
info[i].ldi_file_want_append = lds->lds_want_file_append;
info[i].ldi_file_num_lines_lost = 0;
}
/* Don't open a file in the session directory now -- do that lazily
* so that if there's no output, we don't have an empty file */
return i;
}
static int open_file(int i)
{
int flags;
char *filename;
int n;
/* first check to see if this file is already open
* on someone else's stream - if so, we don't want
* to open it twice
*/
for (n=0; n < OPAL_OUTPUT_MAX_STREAMS; n++) {
if (i == n) {
continue;
}
if (!info[n].ldi_used) {
continue;
}
if (!info[n].ldi_file) {
continue;
}
if (NULL != info[i].ldi_file_suffix &&
NULL != info[n].ldi_file_suffix) {
if (0 != strcmp(info[i].ldi_file_suffix, info[n].ldi_file_suffix)) {
break;
}
}
if (NULL == info[i].ldi_file_suffix &&
NULL != info[n].ldi_file_suffix) {
break;
}
if (NULL != info[i].ldi_file_suffix &&
NULL == info[n].ldi_file_suffix) {
break;
}
if (info[n].ldi_fd < 0) {
break;
}
info[i].ldi_fd = info[n].ldi_fd;
return OPAL_SUCCESS;
}
/* Setup the filename and open flags */
if (NULL != output_dir) {
filename = (char *) malloc(OPAL_PATH_MAX);
if (NULL == filename) {
return OPAL_ERR_OUT_OF_RESOURCE;
}
strncpy(filename, output_dir, OPAL_PATH_MAX);
strcat(filename, "/");
if (NULL != output_prefix) {
strcat(filename, output_prefix);
}
if (info[i].ldi_file_suffix != NULL) {
strcat(filename, info[i].ldi_file_suffix);
} else {
info[i].ldi_file_suffix = NULL;
strcat(filename, "output.txt");
}
flags = O_CREAT | O_RDWR;
if (!info[i].ldi_file_want_append) {
flags |= O_TRUNC;
}
/* Actually open the file */
info[i].ldi_fd = open(filename, flags, 0644);
free(filename); /* release the filename in all cases */
if (-1 == info[i].ldi_fd) {
info[i].ldi_used = false;
return OPAL_ERR_IN_ERRNO;
}
/* Make the file be close-on-exec to prevent child inheritance
* problems */
if (-1 == fcntl(info[i].ldi_fd, F_SETFD, 1)) {
return OPAL_ERR_IN_ERRNO;
}
}
/* Return successfully even if the session dir did not exist yet;
* we'll try opening it later */
return OPAL_SUCCESS;
}
/*
* Free all the resources associated with a descriptor.
*/
static void free_descriptor(int output_id)
{
output_desc_t *ldi;
if (output_id >= 0 && output_id < OPAL_OUTPUT_MAX_STREAMS &&
info[output_id].ldi_used && info[output_id].ldi_enabled) {
ldi = &info[output_id];
if (-1 != ldi->ldi_fd) {
close(ldi->ldi_fd);
}
ldi->ldi_used = false;
/* If we strduped a prefix, suffix, or syslog ident, free it */
if (NULL != ldi->ldi_prefix) {
free(ldi->ldi_prefix);
}
ldi->ldi_prefix = NULL;
if (NULL != ldi->ldi_suffix) {
free(ldi->ldi_suffix);
}
ldi->ldi_suffix = NULL;
if (NULL != ldi->ldi_file_suffix) {
free(ldi->ldi_file_suffix);
}
ldi->ldi_file_suffix = NULL;
if (NULL != ldi->ldi_syslog_ident) {
free(ldi->ldi_syslog_ident);
}
ldi->ldi_syslog_ident = NULL;
}
}
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
static int make_string(char **no_newline_string, output_desc_t *ldi,
const char *format, va_list arglist)
{
size_t len, total_len;
bool want_newline = false;
/* Make the formatted string */
vasprintf(no_newline_string, format, arglist);
total_len = len = strlen(*no_newline_string);
if ('\n' != (*no_newline_string)[len - 1]) {
want_newline = true;
++total_len;
} else if (NULL != ldi->ldi_suffix) {
/* if we have a suffix, then we don't want a
* newline to appear before it
*/
(*no_newline_string)[len - 1] = '\0';
want_newline = true; /* add newline to end after suffix */
/* total_len won't change since we just moved the newline
* to appear after the suffix
*/
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
}
if (NULL != ldi->ldi_prefix) {
total_len += strlen(ldi->ldi_prefix);
}
if (NULL != ldi->ldi_suffix) {
total_len += strlen(ldi->ldi_suffix);
}
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
if (temp_str_len < total_len + want_newline) {
if (NULL != temp_str) {
free(temp_str);
}
temp_str = (char *) malloc(total_len * 2);
if (NULL == temp_str) {
return OPAL_ERR_OUT_OF_RESOURCE;
}
temp_str_len = total_len * 2;
}
if (NULL != ldi->ldi_prefix && NULL != ldi->ldi_suffix) {
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
if (want_newline) {
snprintf(temp_str, temp_str_len, "%s%s%s\n",
ldi->ldi_prefix, *no_newline_string, ldi->ldi_suffix);
} else {
snprintf(temp_str, temp_str_len, "%s%s%s", ldi->ldi_prefix,
*no_newline_string, ldi->ldi_suffix);
}
} else if (NULL != ldi->ldi_prefix) {
if (want_newline) {
snprintf(temp_str, temp_str_len, "%s%s\n",
ldi->ldi_prefix, *no_newline_string);
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
} else {
snprintf(temp_str, temp_str_len, "%s%s", ldi->ldi_prefix,
*no_newline_string);
}
} else if (NULL != ldi->ldi_suffix) {
if (want_newline) {
snprintf(temp_str, temp_str_len, "%s%s\n",
*no_newline_string, ldi->ldi_suffix);
} else {
snprintf(temp_str, temp_str_len, "%s%s",
*no_newline_string, ldi->ldi_suffix);
}
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
} else {
if (want_newline) {
snprintf(temp_str, temp_str_len, "%s\n", *no_newline_string);
} else {
snprintf(temp_str, temp_str_len, "%s", *no_newline_string);
}
}
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
return OPAL_SUCCESS;
}
/*
* Do the actual output. Take a va_list so that we can be called from
* multiple different places, even functions that took "..." as input
* arguments.
*/
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
static int output(int output_id, const char *format, va_list arglist)
{
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
int rc = OPAL_SUCCESS;
char *str, *out = NULL;
output_desc_t *ldi;
/* Setup */
if (!initialized) {
opal_output_init();
}
/* If it's valid, used, and enabled, output */
if (output_id >= 0 && output_id < OPAL_OUTPUT_MAX_STREAMS &&
info[output_id].ldi_used && info[output_id].ldi_enabled) {
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
OPAL_THREAD_LOCK(&mutex);
ldi = &info[output_id];
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
/* Make the strings */
if (OPAL_SUCCESS != (rc = make_string(&str, ldi, format, arglist))) {
OPAL_THREAD_UNLOCK(&mutex);
return rc;
}
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
/* Syslog output -- does not use the newline-appended string */
#if defined(HAVE_SYSLOG)
if (ldi->ldi_syslog) {
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
syslog(ldi->ldi_syslog_priority, "%s", str);
}
#endif
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
/* All others (stdout, stderr, file) use temp_str, potentially
with a newline appended */
out = temp_str;
/* stdout output */
if (ldi->ldi_stdout) {
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
write(fileno(stdout), out, (int)strlen(out));
fflush(stdout);
}
/* stderr output */
if (ldi->ldi_stderr) {
write((-1 == default_stderr_fd) ?
fileno(stderr) : default_stderr_fd,
out, (int)strlen(out));
fflush(stderr);
}
/* File output -- first check to see if the file opening was
* delayed. If so, try to open it. If we failed to open it,
* then just discard (there are big warnings in the
* opal_output.h docs about this!). */
if (ldi->ldi_file) {
if (ldi->ldi_fd == -1) {
if (OPAL_SUCCESS != open_file(output_id)) {
++ldi->ldi_file_num_lines_lost;
} else if (ldi->ldi_file_num_lines_lost > 0) {
char buffer[BUFSIZ];
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
char *out = buffer;
memset(buffer, 0, BUFSIZ);
snprintf(buffer, BUFSIZ - 1,
"[WARNING: %d lines lost because the Open MPI process session directory did\n not exist when opal_output() was invoked]\n",
ldi->ldi_file_num_lines_lost);
write(ldi->ldi_fd, buffer, (int)strlen(buffer));
ldi->ldi_file_num_lines_lost = 0;
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
if (out != buffer) {
free(out);
}
}
}
if (ldi->ldi_fd != -1) {
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
write(ldi->ldi_fd, out, (int)strlen(out));
}
}
OPAL_THREAD_UNLOCK(&mutex);
free(str);
}
This commit represents a bunch of work on a Mercurial side branch. As such, the commit message back to the master SVN repository is fairly long. = ORTE Job-Level Output Messages = Add two new interfaces that should be used for all new code throughout the ORTE and OMPI layers (we already make the search-and-replace on the existing ORTE / OMPI layers): * orte_output(): (and corresponding friends ORTE_OUTPUT, orte_output_verbose, etc.) This function sends the output directly to the HNP for processing as part of a job-specific output channel. It supports all the same outputs as opal_output() (syslog, file, stdout, stderr), but for stdout/stderr, the output is sent to the HNP for processing and output. More on this below. * orte_show_help(): This function is a drop-in-replacement for opal_show_help(), with two differences in functionality: 1. the rendered text help message output is sent to the HNP for display (rather than outputting directly into the process' stderr stream) 1. the HNP detects duplicate help messages and does not display them (so that you don't see the same error message N times, once from each of your N MPI processes); instead, it counts "new" instances of the help message and displays a message every ~5 seconds when there are new ones ("I got X new copies of the help message...") opal_show_help and opal_output still exist, but they only output in the current process. The intent for the new orte_* functions is that they can apply job-level intelligence to the output. As such, we recommend that all new ORTE and OMPI code use the new orte_* functions, not thei opal_* functions. === New code === For ORTE and OMPI programmers, here's what you need to do differently in new code: * Do not include opal/util/show_help.h or opal/util/output.h. Instead, include orte/util/output.h (this one header file has declarations for both the orte_output() series of functions and orte_show_help()). * Effectively s/opal_output/orte_output/gi throughout your code. Note that orte_output_open() takes a slightly different argument list (as a way to pass data to the filtering stream -- see below), so you if explicitly call opal_output_open(), you'll need to slightly adapt to the new signature of orte_output_open(). * Literally s/opal_show_help/orte_show_help/. The function signature is identical. === Notes === * orte_output'ing to stream 0 will do similar to what opal_output'ing did, so leaving a hard-coded "0" as the first argument is safe. * For systems that do not use ORTE's RML or the HNP, the effect of orte_output_* and orte_show_help will be identical to their opal counterparts (the additional information passed to orte_output_open() will be lost!). Indeed, the orte_* functions simply become trivial wrappers to their opal_* counterparts. Note that we have not tested this; the code is simple but it is quite possible that we mucked something up. = Filter Framework = Messages sent view the new orte_* functions described above and messages output via the IOF on the HNP will now optionally be passed through a new "filter" framework before being output to stdout/stderr. The "filter" OPAL MCA framework is intended to allow preprocessing to messages before they are sent to their final destinations. The first component that was written in the filter framework was to create an XML stream, segregating all the messages into different XML tags, etc. This will allow 3rd party tools to read the stdout/stderr from the HNP and be able to know exactly what each text message is (e.g., a help message, another OMPI infrastructure message, stdout from the user process, stderr from the user process, etc.). Filtering is not active by default. Filter components must be specifically requested, such as: {{{ $ mpirun --mca filter xml ... }}} There can only be one filter component active. = New MCA Parameters = The new functionality described above introduces two new MCA parameters: * '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that help messages will be aggregated, as described above. If set to 0, all help messages will be displayed, even if they are duplicates (i.e., the original behavior). * '''orte_base_show_output_recursions''': An MCA parameter to help debug one of the known issues, described below. It is likely that this MCA parameter will disappear before v1.3 final. = Known Issues = * The XML filter component is not complete. The current output from this component is preliminary and not real XML. A bit more work needs to be done to configure.m4 search for an appropriate XML library/link it in/use it at run time. * There are possible recursion loops in the orte_output() and orte_show_help() functions -- e.g., if RML send calls orte_output() or orte_show_help(). We have some ideas how to fix these, but figured that it was ok to commit before feature freeze with known issues. The code currently contains sub-optimal workarounds so that this will not be a problem, but it would be good to actually solve the problem rather than have hackish workarounds before v1.3 final. This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
return rc;
}
int opal_output_get_verbosity(int output_id)
{
if (output_id >= 0 && output_id < OPAL_OUTPUT_MAX_STREAMS && info[output_id].ldi_used) {
return info[output_id].ldi_verbose_level;
} else {
return -1;
}
}