1
1
openmpi/opal/util/output.h

558 строки
21 KiB
C
Исходник Обычный вид История

/*
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2006 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
* University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved.
* Copyright (c) 2007-2008 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
/** @file
* OPAL output stream facility.
*
* The OPAL output stream facility is used to send output from the OPAL
* libraries to output devices. It is meant to fully replace all
* forms of printf() (and friends). Output streams are opened via the
* opal_output_open() function call, and then sent output via
* opal_output_verbose(), OPAL_OUTPUT(), and opal_output(). Streams are
* closed with opal_output_close().
*
* Streams can multiplex output to several kinds of outputs (one of
* each):
*
* - the syslog (if available)
* - standard output
* - standard error
* - file
*
* Which outputs to use are specified during opal_output_open().
*
* WARNING: When using "file" as an output destination, be aware that
* the file may not exist until the session directory for the process
* exists. This is at least part of the way through MPI_INIT (for
* example). Most MCA components and internals of Open MPI won't be
* affected by this, but some RTE / startup aspects of Open MPI will
* not be able to write to a file for output. See opal_output() for
* details on what happens in these cases.
*
* opal_output_open() returns an integer handle that is used in
* successive calls to OPAL_OUTPUT() and opal_output() to send output to
* the stream.
*
* The default "verbose" stream is opened after invoking
* opal_output_init() (and closed after invoking
* opal_output_finalize()). This stream outputs to stderr only, and
* has a stream handle ID of 0.
*
* It is erroneous to have one thread close a stream and have another
* try to write to it. Multiple threads writing to a single stream
* will be serialized in an unspecified order.
*/
#ifndef OPAL_OUTPUT_H_
#define OPAL_OUTPUT_H_
#include "opal_config.h"
#ifdef HAVE_STDARG_H
#include <stdarg.h>
#endif
#include "opal/class/opal_object.h"
BEGIN_C_DECLS
/* There are systems where all output needs to be redirected to syslog
* and away from stdout/stderr or files - e.g., embedded systems whose
* sole file system is in flash. To support such systems, we provide
* the following environmental variables that support redirecting -all-
* output (both from opal_output and stdout/stderr of processes) to
* syslog:
*
* OPAL_OUTPUT_REDIRECT - set to "syslog" to redirect to syslog. Other
* options may someday be supported
* OPAL_OUTPUT_SYSLOG_PRI - set to "info", "error", or "warn" to have
* output sent to syslog at that priority
* OPAL_OUTPUT_SYSLOG_IDENT - a string identifier for the log
*
* We also define two global variables that notify all other
* layers that output is being redirected to syslog at the given
* priority. These are used, for example, by the IO forwarding
* subsystem to tell it to dump any collected output directly to
* syslog instead of forwarding it to another location.
*/
OPAL_DECLSPEC extern bool opal_output_redirected_to_syslog;
OPAL_DECLSPEC extern int opal_output_redirected_syslog_pri;
/**
* \class opal_output_stream_t
*
* Structure used to request the opening of a OPAL output stream. A
* pointer to this structure is passed to opal_output_open() to tell
* the opal_output subsystem where to send output for a given stream.
* It is valid to specify multiple destinations of output for a stream
* -- output streams can be multiplexed to multiple different
* destinations through the opal_output facility.
*
* Note that all strings in this struct are cached on the stream by
* value; there is no need to keep them allocated after the return
* from opal_output_open().
*/
struct opal_output_stream_t {
/** Class parent */
opal_object_t super;
/**
* Indicate the starting verbosity level of the stream.
*
* Verbose levels are a convenience mechanisms, and are only
* consulted when output is sent to a stream through the
* opal_output_verbose() function. Verbose levels are ignored in
* OPAL_OUTPUT() and opal_output().
*
* Valid verbose levels typically start at 0 (meaning "minimal
* information"). Higher verbosity levels generally indicate that
* more output and diagnostics should be displayed.
*/
int lds_verbose_level;
/**
* When opal_output_stream_t::lds_want_syslog is true, this field is
* examined to see what priority output from the stream should be
* sent to the syslog.
*
* This value should be set as per the syslog(3) man page. It is
* typically the OR value of "facilty" and "level" values described
* in the man page.
*/
int lds_syslog_priority;
/**
* When opal_output_stream_t::lds_want_syslog is true, this field is
* examined to see what ident value should be passed to openlog(3).
*
* If a NULL value is given, the string "opal" is used.
*/
#if !defined(__WINDOWS__)
char *lds_syslog_ident;
#else
HANDLE lds_syslog_ident;
#endif /* !defined(__WINDOWS__) */
/**
* String prefix added to all output on the stream.
*
* When this field is non-NULL, it is prefixed to all lines of
* output on the stream. When this field is NULL, no prefix is
* added to each line of output in the stream. The prefix is copied
* to an internal structure in the call to opal_output_open()!
*/
char *lds_prefix;
/**
* String suffix added to all output on the stream.
*
* When this field is non-NULL, it is appended to all lines of
* output on the stream. When this field is NULL, no suffix is
* added to each line of output in the stream. The suffix is copied
* to an internal structure in the call to opal_output_open()!
*/
char *lds_suffix;
/**
* Indicates whether the output of the stream is
* debugging/developer-only output or not.
*
* This field should be "true" if the output is for debugging
* purposes only. In that case, the output will never be sent to
* the stream unless OPAL was configured with --enable-debug.
*/
bool lds_is_debugging;
/**
* Indicates whether output of the stream should be sent to the
* syslog or not.
*
* If this field is true, output from this stream is sent to the
* syslog, and the following fields are also examined:
*
* - lds_syslog_priority
* - lds_syslog_ident
* - lds_prefix
*
* If this field is false, the above three fields are ignored.
*/
bool lds_want_syslog;
/**
* Whether to send stream output to stdout or not.
*
* If this field is true, stream output is sent to stdout.
*/
bool lds_want_stdout;
/**
* Whether to send stream output to stderr or not.
*
* If this field is true, stream output is sent to stderr.
*/
bool lds_want_stderr;
/**
* Whether to send stream output to a file or not.
*
* When this field is true, stream output is sent to a file, and the
* following fields are also examined:
*
* - lds_want_file_append
* - lda_file_suffix
*/
bool lds_want_file;
/**
* When opal_output_stream_t::lds_want_file is true, this field
* indicates whether to append the file (if it exists) or overwrite
* it.
*
* If false, the file is opened with the O_TRUNC flag.
*/
bool lds_want_file_append;
/**
* When opal_output_stream_t::lds_want_file is true, this field
* indicates the string suffix to add to the filename.
*
* The output file will be in the directory and begin with the
* prefix set by opal_output_set_output_file_info() (e.g.,
* "$dir/$prefix$suffix"). If this field is NULL and
* lds_want_file is true, then the suffix "output.txt" is used.
*
* Note that it is possible that the output directory may not
* exist when opal_output_open() is invoked. See opal_output()
* for details on what happens in this situation.
*/
char *lds_file_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
};
/**
* Convenience typedef
*/
typedef struct opal_output_stream_t opal_output_stream_t;
/**
* Initializes the output stream system and opens a default
* "verbose" stream.
*
* @retval true Upon success.
* @retval false Upon failure.
*
* This should be the first function invoked in the output
* subsystem. After this call, the default "verbose" stream is open
* and can be written to via calls to opal_output_verbose() and
* opal_output_error().
*
* By definition, the default verbose stream has a handle ID of 0,
* and has a verbose level of 0.
*/
OPAL_DECLSPEC bool opal_output_init(void);
/**
* Shut down the output stream system.
*
* Shut down the output stream system, including the default verbose
* stream.
*/
OPAL_DECLSPEC void opal_output_finalize(void);
/**
* Opens an output stream.
*
* @param lds A pointer to opal_output_stream_t describing what the
* characteristics of the output stream should be.
*
* This function opens an output stream and returns an integer
* handle. The caller is responsible for maintaining the handle and
* using it in successive calls to OPAL_OUTPUT(), opal_output(),
* opal_output_switch(), and opal_output_close().
*
* If lds is NULL, the default descriptions will be used, meaning
* that output will only be sent to stderr.
*
* It is safe to have multiple threads invoke this function
* simultaneously; their execution will be serialized in an
* unspecified manner.
*
* Be sure to see opal_output() for a description of what happens
* when open_open() / opal_output() is directed to send output to a
* file but the process session directory does not yet exist.
*/
OPAL_DECLSPEC int opal_output_open(opal_output_stream_t *lds);
/**
* Re-opens / redirects an output stream.
*
* @param output_id Stream handle to reopen
* @param lds A pointer to opal_output_stream_t describing what the
* characteristics of the reopened output stream should be.
*
* This function redirects an existing stream into a new [set of]
* location[s], as specified by the lds parameter. If the output_is
* passed is invalid, this call is effectively the same as opening a
* new stream with a specific stream handle.
*/
OPAL_DECLSPEC int opal_output_reopen(int output_id, opal_output_stream_t *lds);
/**
* Enables and disables output streams.
*
* @param output_id Stream handle to switch
* @param enable Boolean indicating whether to enable the stream
* output or not.
*
* @returns The previous enable state of the stream (true == enabled,
* false == disabled).
*
* The output of a stream can be temporarily disabled by passing an
* enable value to false, and later resumed by passing an enable
* value of true. This does not close the stream -- it simply tells
* the opal_output subsystem to intercept and discard any output sent
* to the stream via OPAL_OUTPUT() or opal_output() until the output
* is re-enabled.
*/
OPAL_DECLSPEC bool opal_output_switch(int output_id, bool enable);
/**
* \internal
*
* Reopens all existing output streams.
*
* This function should never be called by user applications; it is
* typically only invoked after a restart (i.e., in a new process)
* where output streams need to be re-initialized.
*/
OPAL_DECLSPEC void opal_output_reopen_all(void);
/**
* Close an output stream.
*
* @param output_id Handle of the stream to close.
*
* Close an output stream. No output will be sent to the stream
* after it is closed. Be aware that output handles tend to be
* re-used; it is possible that after a stream is closed, if another
* stream is opened, it will get the same handle value.
*/
OPAL_DECLSPEC void opal_output_close(int output_id);
/**
* Main function to send output to a stream.
*
* @param output_id Stream id returned from opal_output_open().
* @param format printf-style format string.
* @param varargs printf-style varargs list to fill the string
* specified by the format parameter.
*
* This is the main function to send output to custom streams (note
* that output to the default "verbose" stream is handled through
* opal_output_verbose() and opal_output_error()).
*
* It is never necessary to send a trailing "\n" in the strings to
* this function; some streams requires newlines, others do not --
* this function will append newlines as necessary.
*
* Verbosity levels are ignored in this function.
*
* Note that for output streams that are directed to files, the
* files are stored under the process' session directory. If the
* session directory does not exist when opal_output() is invoked,
* the output will be discarded! Once the session directory is
* created, opal_output() will automatically create the file and
* writing to it.
*/
OPAL_DECLSPEC void opal_output(int output_id, const char *format, ...) __opal_attribute_format__(__printf__, 2, 3);
/**
* Send output to a stream only if the passed verbosity level is
* high enough.
*
* @param output_id Stream id returned from opal_output_open().
* @param level Target verbosity level.
* @param format printf-style format string.
* @param varargs printf-style varargs list to fill the string
* specified by the format parameter.
*
* Output is only sent to the stream if the current verbosity level
* is greater than or equal to the level parameter. This mechanism
* can be used to send "information" kinds of output to user
* applications, but only when the user has asked for a high enough
* verbosity level.
*
* It is never necessary to send a trailing "\n" in the strings to
* this function; some streams requires newlines, others do not --
* this function will append newlines as necessary.
*
* This function is really a convenience wrapper around checking the
* current verbosity level set on the stream, and if the passed
* level is less than or equal to the stream's verbosity level, this
* function will effectively invoke opal_output to send the output to
* the stream.
*
* @see opal_output_set_verbosity()
*/
OPAL_DECLSPEC void opal_output_verbose(int verbose_level, int output_id,
const char *format, ...) __opal_attribute_format__(__printf__, 3, 4);
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
/**
* Same as opal_output_verbose(), but takes a va_list form of varargs.
*/
OPAL_DECLSPEC void opal_output_vverbose(int verbose_level, int output_id,
const char *format, va_list ap) __opal_attribute_format__(__printf__, 3, 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
/**
* Send output to a string if the verbosity level is high enough.
*
* @param output_id Stream id returned from opal_output_open().
* @param level Target verbosity level.
* @param format printf-style format string.
* @param varargs printf-style varargs list to fill the string
* specified by the format parameter.
*
* Exactly the same as opal_output_verbose(), except the output it
* sent to a string instead of to the stream. If the verbose
* level is not high enough, NULL is returned. The caller is
* responsible for free()'ing the returned string.
*/
OPAL_DECLSPEC char *opal_output_string(int verbose_level, int output_id,
const char *format, ...) __opal_attribute_format__(__printf__, 3, 4);
/**
* Same as opal_output_string, but accepts a va_list form of varargs.
*/
OPAL_DECLSPEC char *opal_output_vstring(int verbose_level, int output_id,
const char *format, va_list ap) __opal_attribute_format__(__printf__, 3, 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
/**
* Set the verbosity level for a stream.
*
* @param output_id Stream id returned from opal_output_open().
* @param level New verbosity level
*
* This function sets the verbosity level on a given stream. It
* will be used for all future invocations of opal_output_verbose().
*/
OPAL_DECLSPEC void opal_output_set_verbosity(int output_id, int level);
/**
* Get the verbosity level for a stream
*
* @param output_id Stream id returned from opal_output_open()
* @returns Verbosity of stream
*/
OPAL_DECLSPEC int opal_output_get_verbosity(int output_id);
/**
* Set characteristics for output files.
*
* @param dir Directory where output files will go
* @param olddir If non-NULL, the directory where output files
* were previously opened
* @param prefix Prefix of files in the output directory
* @param oldprefix If non-NULL, the old prefix
*
* This function controls the final filename used for all new
* output streams that request output files. Specifically, when
* opal_output_stream_t::lds_want_file is true, the output
* filename will be of the form $dir/$prefix$suffix.
*
* The default value for the output directory is whatever is
* specified in the TMPDIR environment variable if it exists, or
* $HOME if it does not. The default value for the prefix is
* "output-pid<pid>-" (where "<pid>" is replaced by the PID of the
* current process).
*
* If dir or prefix are NULL, new values are not set. The strings
* represented by dir and prefix are copied into internal storage;
* it is safe to pass string constants or free() these values
* after opal_output_set_output_file_info() returns.
*
* If olddir or oldprefix are not NULL, copies of the old
* directory and prefix (respectively) are returned in these
* parameters. The caller is responsible for calling (free) on
* these values. This allows one to get the old values, output an
* output file in a specific directory and/or with a specific
* prefix, and then restore the old values.
*
* Note that this function only affects the creation of \em new
* streams -- streams that have already started writing to output
* files are not affected (i.e., their output files are not moved
* to the new directory). More specifically, the opal_output
* system only opens/creates output files lazily -- so calling
* this function affects both new streams \em and any stream that
* was previously opened but had not yet output anything.
*/
OPAL_DECLSPEC void opal_output_set_output_file_info(const char *dir,
const char *prefix,
char **olddir,
char **oldprefix);
#if OPAL_ENABLE_DEBUG
/**
* Main macro for use in sending debugging output to output streams;
* will be "compiled out" when OPAL is configured without
* --enable-debug.
*
* @see opal_output()
*/
#define OPAL_OUTPUT(a) opal_output a
/**
* Macro for use in sending debugging output to the output
* streams. Will be "compiled out" when OPAL is configured
* without --enable-debug.
*
* @see opal_output_verbose()
*/
#define OPAL_OUTPUT_VERBOSE(a) opal_output_verbose a
#else
/**
* Main macro for use in sending debugging output to output streams;
* will be "compiled out" when OPAL is configured without
* --enable-debug.
*
* @see opal_output()
*/
#define OPAL_OUTPUT(a)
/**
* Macro for use in sending debugging output to the output
* streams. Will be "compiled out" when OPAL is configured
* without --enable-debug.
*
* @see opal_output_verbose()
*/
#define OPAL_OUTPUT_VERBOSE(a)
#endif
/**
* Declare the class of this type. Note that the constructor for
* this class is for convenience only -- it is \em not necessary
* to be invoked. If the constructor it used, it sets all values
* in the struct to be false / 0 (i.e., turning off all output).
* The intended usage is to invoke the constructor and then enable
* the output fields that you want.
*/
OPAL_DECLSPEC OBJ_CLASS_DECLARATION(opal_output_stream_t);
END_C_DECLS
#endif /* OPAL_OUTPUT_H_ */