libiberty:

	* pex-common.c: New file.
	* pex-one.c: New file.
	* pexecute.c: New file.
	* pex-common.h: Include <stdio.h>.
	(struct pex_obj): Define.
	(struct pex_funcs): Define.
	(pex_init_common): Declare.
	* pex-unix.c: Rewrite.
	* pex-win32.c: Rewrite.
	* pex-djgpp.c: Rewrite.
	* pex-msdos.c: Rewrite.
	* testsuite/text-pexecute.c: New file.
	* pexecute.txh: Rewrite.
	* configure.ac: Check for wait3 and wait4.  Set CHECK to
	really-check rather than check-cplus-dem.
	* functions.texi: Rebuild.
	* Makefile.in: Rebuild dependencies.
	(CFILES): Add pexecute.c, pex-common.c, pex-one.c.
	(REQUIRED_OFILES): Add pexecute.o, pex-common.o, pex-one.o.
	* testsuite/Makefile.in (really-check): New target.
	(check-pexecute, test-pexecute): New targets.
	* configure: Rebuild.
include:
	* libiberty.h: Include <stdio.h>.
	(PEX_RECORD_TIMES, PEX_USE_PIPES, PEX_SAVE_TEMPS): Define.
	(PEX_LAST, PEX_SEARCH, PEX_SUFFIX, PEX_STDERR_TO_STDOUT): Define.
	(PEX_BINARY_INPUT, PEX_BINARY_OUTPUT): Define.
	(pex_init, pex_run, pex_read_output): Declare.
	(pex_get_status, pex_get_times, pex_free, pex_one): Declare.
	(struct pex_time): Define.

1 files changed, 165 insertions, 43 deletions

diff --git a/libiberty/pexecute.txh b/libiberty/pexecute.txh
index 269f031..c06d55c 100644
--- a/libiberty/pexecute.txh
+++ b/libiberty/pexecute.txh
@@ -1,63 +1,185 @@
-@deftypefn Extension int pexecute (const char *@var{program}, char * const *@var{argv}, const char *@var{this_pname}, const char *@var{temp_base}, char **@var{errmsg_fmt}, char **@var{errmsg_arg}, int flags)
+@deftypefn Extension struct pex_obj *pex_init (int @var{flags}, const char *@var{pname}, const char *@var{tempbase})
 
-Executes a program.
+Prepare to execute one or more programs, with standard output of each
+program fed to standard input of the next.  This is a system
+independent interface to execute a pipeline.
 
-@var{program} and @var{argv} are the arguments to
-@code{execv}/@code{execvp}.
+@var{flags} is a bitwise combination of the following:
 
-@var{this_pname} is name of the calling program (i.e., @code{argv[0]}).
+@table @code
 
-@var{temp_base} is the path name, sans suffix, of a temporary file to
-use if needed.  This is currently only needed for MS-DOS ports that
-don't use @code{go32} (do any still exist?).  Ports that don't need it
-can pass @code{NULL}.
+@vindex PEX_RECORD_TIMES
+@item PEX_RECORD_TIMES
+Record subprocess times if possible.
 
-(@code{@var{flags} & PEXECUTE_SEARCH}) is non-zero if @env{PATH}
-should be searched (??? It's not clear that GCC passes this flag
-correctly).  (@code{@var{flags} & PEXECUTE_FIRST}) is nonzero for the
-first process in chain.  (@code{@var{flags} & PEXECUTE_FIRST}) is
-nonzero for the last process in chain.  The first/last flags could be
-simplified to only mark the last of a chain of processes but that
-requires the caller to always mark the last one (and not give up
-early if some error occurs).  It's more robust to require the caller
-to mark both ends of the chain.
+@vindex PEX_USE_PIPES
+@item PEX_USE_PIPES
+Use pipes for communication between processes, if possible.
 
-The result is the pid on systems like Unix where we
-@code{fork}/@code{exec} and on systems like WIN32 and OS/2 where we
-use @code{spawn}.  It is up to the caller to wait for the child.
+@vindex PEX_SAVE_TEMPS
+@item PEX_SAVE_TEMPS
+Don't delete temporary files used for communication between
+processes.
 
-The result is the @code{WEXITSTATUS} on systems like MS-DOS where we
-@code{spawn} and wait for the child here.
+@end table
 
-Upon failure, @var{errmsg_fmt} and @var{errmsg_arg} are set to the
-text of the error message with an optional argument (if not needed,
-@var{errmsg_arg} is set to @code{NULL}), and @minus{}1 is returned.
-@code{errno} is available to the caller to use.
+@var{pname} is the name of program to be executed, used in error
+messages.  @var{tempbase} is a base name to use for any required
+temporary files; it may be @code{NULL} to use a randomly chosen name.
 
 @end deftypefn
 
-@deftypefn Extension int pwait (int @var{pid}, int *@var{status}, int @var{flags})
+@deftypefn Extension const char *pex_run (struct pex_obj *@var{obj}, int @var{flags}, const char *@var{executable}, char * const *@var{argv}, const char *@var{outname}, const char *@var{errname}, int *@var{err})
+
+Execute one program in a pipeline.  On success this returns
+@code{NULL}.  On failure it returns an error message, a statically
+allocated string.
+
+@var{obj} is returned by a previous call to @code{pex_init}.
+
+@var{flags} is a bitwise combination of the following:
+
+@table @code
+
+@vindex PEX_LAST
+@item PEX_LAST
+This must be set on the last program in the pipeline.  In particular,
+it should be set when executing a single program.  The standard output
+of the program will be sent to @var{outname}, or, if @var{outname} is
+@code{NULL}, to the standard output of the calling program.  This
+should not be set if you want to call @code{pex_read_output}
+(described below).  After a call to @code{pex_run} with this bit set,
+@var{pex_run} may no longer be called with the same @var{obj}.
+
+@vindex PEX_SEARCH
+@item PEX_SEARCH
+Search for the program using the user's executable search path.
+
+@vindex PEX_SUFFIX
+@item PEX_SUFFIX
+@var{outname} is a suffix.  See the description of @var{outname},
+below.
+
+@vindex PEX_STDERR_TO_STDOUT
+@item PEX_STDERR_TO_STDOUT
+Send the program's standard error to standard output, if possible.
+
+@vindex PEX_BINARY_INPUT
+@vindex PEX_BINARY_OUTPUT
+@item PEX_BINARY_INPUT
+@itemx PEX_BINARY_OUTPUT
+The standard input (output) of the program should be read (written) in
+binary mode rather than text mode.  These flags are ignored on systems
+which do not distinguish binary mode and text mode, such as Unix.  For
+proper behavior these flags should match appropriately--a call to
+@code{pex_run} using @code{PEX_BINARY_OUTPUT} should be followed by a
+call using @code{PEX_BINARY_INPUT}.
+@end table
+
+@var{executable} is the program to execute.  @var{argv} is the set of
+arguments to pass to the program; normally @code{@var{argv}[0]} will
+be a copy of @var{executable}.
+
+@var{outname} is used to set the name of the file to use for standard
+output.  There are two cases in which no output file will be used: 1)
+if @code{PEX_LAST} is not set in @var{flags}, and @code{PEX_USE_PIPES}
+was set in the call to @code{pex_init}, and the system supports pipes;
+2) if @code{PEX_LAST} is set in @var{flags}, and @var{outname} is
+@code{NULL}.  Otherwise the code will use a file to hold standard
+output.  If @code{PEX_LAST} is not set, this file is considered to be
+a temporary file, and it will be removed when no longer needed, unless
+@code{PEX_SAVE_TEMPS} was set in the call to @code{pex_init}.
+
+There are two cases to consider when setting the name of the file to
+hold standard output.
+
+First case: @code{PEX_SUFFIX} is set in @var{flags}.  In this case
+@var{outname} may not be @code{NULL}.  If the @var{tempbase} parameter
+to @code{pex_init} was not @code{NULL}, then the output file name is
+the concatenation of @var{tempbase} and @var{outname}.  If
+@var{tempbase} was @code{NULL}, then the output file name is a random
+file name ending in @var{outname}.
+
+Second case: @code{PEX_SUFFIX} was not set in @var{flags}.  In this
+case, if @var{outname} is not @code{NULL}, it is used as the output
+file name.  If @var{outname} is @code{NULL}, and @var{tempbase} was
+not NULL, the output file name is randomly chosen using
+@var{tempbase}.  Otherwise the output file name is chosen completely
+at random.
+
+@var{errname} is the file name to use for standard error output.  If
+it is @code{NULL}, standard error is the same as the caller.
+Otherwise, standard error is written to the named file.
+
+On an error return, the code sets @code{*@var{err}} to an @code{errno}
+value, or to 0 if there is no relevant @code{errno}.
+
+@end deftypefn
+
+@deftypefn Extension FILE * pex_read_output (struct pex_obj *@var{obj}, int @var{binary})
+
+Returns a @code{FILE} pointer which may be used to read the standard
+output of the last program in the pipeline.  When this is used,
+@code{PEX_LAST} should not be used in a call to @code{pex_run}.  After
+this is called, @code{pex_run} may no longer be called with the same
+@var{obj}.  @var{binary} should be non-zero if the file should be
+opened in binary mode.  Don't call @code{fclose} on the returned file;
+it will be closed by @code{pex_free}.
+
+@end deftypefn
+
+@deftypefn Extension int pex_get_status (struct pex_obj *@var{obj}, int @var{count}, int *@var{vector})
+
+Returns the exit status of all programs run using @var{obj}.
+@var{count} is the number of results expected.  The results will be
+placed into @var{vector}.  The results are in the order of the calls
+to @code{pex_run}.  Returns 0 on error, 1 on success.
+
+@end deftypefn
+
+@deftypefn Extension int pex_get_times (struct pex_obj *@var{obj}, int @var{count}, struct pex_time *@var{vector})
 
-Waits for a program started by @code{pexecute} to finish.
+Returns the process execution times of all programs run using
+@var{obj}.  @var{count} is the number of results expected.  The
+results will be placed into @var{vector}.  The results are in the
+order of the calls to @code{pex_run}.  Returns 0 on error, 1 on
+success.
 
-@var{pid} is the process id of the task to wait for. @var{status} is
-the `status' argument to wait. @var{flags} is currently unused
-(allows future enhancement without breaking upward compatibility).
-Pass 0 for now.
+@code{struct pex_time} has the following fields: @code{user_seconds},
+@code{user_microseconds}, @code{system_seconds},
+@code{system_microseconds}.  On systems which do not support reporting
+process times, all the fields will be set to @code{0}.
 
-The result is the pid of the child reaped, or -1 for failure
-(@code{errno} says why).
+@end deftypefn
+
+@deftypefn Extension void pex_free (struct pex_obj @var{obj})
 
-On systems that don't support waiting for a particular child,
-@var{pid} is ignored.  On systems like MS-DOS that don't really
-multitask @code{pwait} is just a mechanism to provide a consistent
-interface for the caller.
+Clean up and free all data associated with @var{obj}.
 
 @end deftypefn
 
-@undocumented pfinish
+@deftypefn Extension const char *pex_one (int @var{flags}, const char *@var{executable}, char * const *@var{argv}, const char *@var{pname}, const char *@var{outname}, const char *@var{errname}, int *@var{status}, int *@var{err})
+
+An interface to @code{pex_init} to permit the easy execution of a
+single program.  The return value and most of the parameters are as
+for a call to @code{pex_run}.  @var{flags} is restricted to a
+combination of @code{PEX_SEARCH}, @code{PEX_STDERR_TO_STDOUT}, and
+@code{PEX_BINARY_OUTPUT}.  @var{outname} is interpreted as if
+@code{PEX_LAST} were set.  On a successful return, *@var{status} will
+be set to the exit status of the program.
+
+@end deftypefn
+
+@deftypefn Extension int pexecute (const char *@var{program}, char * const *@var{argv}, const char *@var{this_pname}, const char *@var{temp_base}, char **@var{errmsg_fmt}, char **@var{errmsg_arg}, int flags)
 
-pfinish: finish generation of script
+This is the old interface to execute one or more programs.  It is
+still supported for compatibility purposes, but is no longer
+documented.
 
-pfinish is necessary for systems like MPW where a script is generated
-that runs the requested programs.
+@end deftypefn
+
+@deftypefn Extension int pwait (int @var{pid}, int *@var{status}, int @var{flags})
+
+Another part of the old execution interface.
+
+@end deftypefn