This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH v3 5/6] Share fork_inferior et al with gdbserver


On Wednesday, February 15 2017, Pedro Alves wrote:

> Hi Sergio,
>
> I think I'll likely have more comments, but this large, so
> here's a first pass.

Cool, thanks for the comments so far.  I had to spend a bit of time
thinking about some of them and addressing everything.  Sorry about the
delay.

>> diff --git a/gdb/common/common-fork-child.c b/gdb/common/common-fork-child.c
>> new file mode 100644
>> index 0000000..f4d7866
>> --- /dev/null
>> +++ b/gdb/common/common-fork-child.c
>> @@ -0,0 +1,593 @@
>> +/* Fork a Unix child process, and set up to debug it, for GDB and GDBserver.
>> +
>> +   Copyright (C) 1990-2017 Free Software Foundation, Inc.
>> +
>> +   Originally contributed by Cygnus Support.
>
> We should just take the opportunity to drop that per recent policy.
> Plenty of files added by Cygnus (nowadays Red Hat) don't have that marker,
> this won't be missed, IMO.

Done.

>> +
>> +   This file is part of GDB.
>> +
>> +   This program is free software; you can redistribute it and/or modify
>> +   it under the terms of the GNU General Public License as published by
>> +   the Free Software Foundation; either version 3 of the License, or
>> +   (at your option) any later version.
>> +
>> +   This program is distributed in the hope that it will be useful,
>> +   but WITHOUT ANY WARRANTY; without even the implied warranty of
>> +   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>> +   GNU General Public License for more details.
>> +
>> +   You should have received a copy of the GNU General Public License
>> +   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
>> +
>
>> +int
>> +fork_inferior (char *exec_file_arg, char *allargs, char **env,
>> +	       void (*traceme_fun) (void), void (*init_trace_fun) (int),
>> +	       void (*pre_trace_fun) (void), char *shell_file_arg,
>> +	       void (*exec_fun)(const char *file, char * const *argv,
>> +				char * const *env))
>> +{
>
>> +  /* Retain a copy of our environment variables, since the child will
>> +     replace the value of environ and if we're vforked, we have to
>> +     restore it.  */
>> +  save_our_env = environ;
>> +
>> +  /* Likewise the current UI.  */
>> +  save_ui = current_ui;
>
> Hmm, making this compile on gdbserver is of course a hack,
> since there's not such thing as a "UI" concept on gdbserver...
>
>> +
>> +  /* Tell the terminal handling subsystem what tty we plan to run on;
>> +     it will just record the information for later.  */
>> +  new_tty_prefork (inferior_io_terminal);
>
> I wonder about calling here instead a more generalized hook, and
> moving the save_ui saving and the new_tty_prefork calls to
> that hook's gdb implementation.  Might be easier to do that after
> the series is in...

Right, it is a hack indeed, and not a beautiful one.

I implemented a bunch of new functions that solve this problem.  Two of
them, tty_{pre,post}fork_hook, are responsible for saving/restoring the
current 'struct ui' and also for calling the new_tty_{pre,post}fork
functions (which are now static).  One more function was needed:
switch_ui_postfork, which is responsible for switching the current_ui to
main_ui, as is done currently after we successfully fork.  These 3
functions are implemented only on GDB; they're stubs on gdbserver.

>> +
>> +  /* It is generally good practice to flush any possible pending stdio
>> +     output prior to doing a fork, to avoid the possibility of both
>> +     the parent and child flushing the same data after the fork.  */
>> +  gdb_flush_out_err ();
>> +
>> +  /* If there's any initialization of the target layers that must
>> +     happen to prepare to handle the child we're about fork, do it
>> +     now...  */
>> +  if (pre_trace_fun != NULL)
>> +    (*pre_trace_fun) ();
>> +
>
>
>> +++ b/gdb/common/common-top.h
>> @@ -0,0 +1,31 @@
>> +/* Common top level stuff for GDB and GDBserver.
>> +
>> +   Copyright (C) 1986-2017 Free Software Foundation, Inc.
>> +
>> +   This file is part of GDB.
>> +
>> +   This program is free software; you can redistribute it and/or modify
>> +   it under the terms of the GNU General Public License as published by
>> +   the Free Software Foundation; either version 3 of the License, or
>> +   (at your option) any later version.
>> +
>> +   This program is distributed in the hope that it will be useful,
>> +   but WITHOUT ANY WARRANTY; without even the implied warranty of
>> +   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>> +   GNU General Public License for more details.
>> +
>> +   You should have received a copy of the GNU General Public License
>> +   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
>> +
>> +#ifndef COMMON_TOP_H
>> +#define COMMON_TOP_H
>> +
>> +/* The main UI.  This is the UI that is bound to stdin/stdout/stderr.
>> +   It always exists and is created automatically when GDB starts
>> +   up.  */
>> +extern struct ui *main_ui;
>> +
>> +/* The current UI.  */
>> +extern struct ui *current_ui;
>> +
>
> Not exactly thrilled with moving this to common.

See comment above.  This is now back on GDB land.

>> +#endif /* ! COMMON_TOP_H */
>
>>  /* Implement the "unset exec-wrapper" command.  */
>> @@ -585,7 +73,7 @@ Show the wrapper for running programs."), NULL,
>>             &unsetlist);
>>  
>>    add_setshow_boolean_cmd ("startup-with-shell", class_support,
>> -			   &startup_with_shell, _("\
>> +			   (int *) &startup_with_shell, _("\
>
> That's invalid C/C++.  Don't do that.

Ops, thanks for pointing that out.  I decided to revert the change and
use an integer instead of a boolean there; at least until we have proper
add_setshow* functions for real booleans.

>>  Set use of shell to start subprocesses.  The default is on."), _("\
>>  Show use of shell to start subprocesses."), NULL,
>>  			   NULL,
>
>
>> --- a/gdb/gdbserver/inferiors.c
>> +++ b/gdb/gdbserver/inferiors.c
>> @@ -29,6 +29,8 @@ struct thread_info *current_thread;
>>  
>>  #define get_thread(inf) ((struct thread_info *)(inf))
>>  
>> +ptid_t inferior_ptid;
>
> What do we need this for?  gdbserver already has
> a "current thread" global.  Another one looks like asking
> for out-of-sync trouble.

This is needed because fork_inferior et al reference this variable
directly, and so I moved the 'extern' declaration of it to commom/.

A possible solution to this would be to create a get/set pair of
functions for it, but I'm not sure this would be a good idea due to (a)
the number of direct references to it, and (b) the fact that these
functions would probably end up being stubs on gdbserver as well.

>>  /* See common/common-gdbthread.h.  */
>>  
>>  void
>>  init_thread_list (void)
>>  {
>> -  /* To be implemented.  */
>> +  /* Placeholder needed for fork_inferior.  No action is needed.  */
>
> Pick one of the comments, and move it to the right patch.

Sorry, fixed.

>>  }
>>  
>
>> -/* Add a process to the common process list, and set its private
>> -   data.  */
>> +/* Update process represented by PID with necessary info.  */
>>  
>>  static struct process_info *
>> -linux_add_process (int pid, int attached)
>> +linux_update_process (int pid)
>
> I'm not sure I understand the need for this yet.  I need
> to look deeper.  "update what?  why?"  Or maybe the
> comments should be improved.  :-)

The reason these 'update' functions were created is because
fork_inferior already creates the process/thread structures, but we (the
caller) still need to fill in some of the fields of these structures
with more information.  They are the same functions that existed before,
but now we work with an existing process/thread, while before we
*created* these structures.

>>  {
>> -  struct process_info *proc;
>> +  struct process_info *proc = find_process_pid (pid);
>>  
>> -  proc = add_process (pid, attached);
>> +  gdb_assert (proc != NULL);
>>    proc->priv = XCNEW (struct process_info_private);
>>  
>>    if (the_low_target.new_process != NULL)
>> @@ -432,6 +433,16 @@ linux_add_process (int pid, int attached)
>>    return proc;
>>  }
>>  
>
>> +/* Update the lwp associated to thread represented by PTID.  */
>> +
>> +static struct lwp_info *
>> +update_thread_lwp (ptid_t ptid)
>
> Ditto.
>
>>  /* Start an inferior process and returns its pid.
>>     ALLARGS is a vector of program-name and args. */
>>  
>>  static int
>> -linux_create_inferior (char *program, char **allargs)
>> +linux_create_inferior (std::vector<char *> &program_argv)
>
> "const std::vector<char *> &" ?  Same comment for the whole
> call chain, of course.

Fixed.

>>  {
>>    struct lwp_info *new_lwp;
>>    int pid;
>>    ptid_t ptid;
>>    struct cleanup *restore_personality
>>      = maybe_disable_address_space_randomization (disable_randomization);
>> +  std::string program_args = stringify_argv (program_argv);
>>  
>> -#if defined(__UCLIBC__) && defined(HAS_NOMMU)
>> -  pid = vfork ();
>> -#else
>> -  pid = fork ();
>> -#endif
>> -  if (pid < 0)
>> -    perror_with_name ("fork");
>> +  pre_fork_inferior (program_argv);
>>  
>> -  if (pid == 0)
>> -    {
>> -      close_most_fds ();
>> -      ptrace (PTRACE_TRACEME, 0, (PTRACE_TYPE_ARG3) 0, (PTRACE_TYPE_ARG4) 0);
>> -
>> -      setpgid (0, 0);
>> -
>> -      /* If gdbserver is connected to gdb via stdio, redirect the inferior's
>> -	 stdout to stderr so that inferior i/o doesn't corrupt the connection.
>> -	 Also, redirect stdin to /dev/null.  */
>> -      if (remote_connection_is_stdio ())
>> -	{
>> -	  close (0);
>> -	  open ("/dev/null", O_RDONLY);
>> -	  dup2 (2, 1);
>> -	  if (write (2, "stdin/stdout redirected\n",
>> -		     sizeof ("stdin/stdout redirected\n") - 1) < 0)
>> -	    {
>> -	      /* Errors ignored.  */;
>> -	    }
>> -	}
>> -
>> -      restore_original_signals_state ();
>> -
>> -      execv (program, allargs);
>> -      if (errno == ENOENT)
>> -	execvp (program, allargs);
>> -
>> -      fprintf (stderr, "Cannot exec %s: %s.\n", program,
>> -	       strerror (errno));
>> -      fflush (stderr);
>> -      _exit (0177);
>> -    }
>> +  pid = fork_inferior (program_argv[0],
>> +		       (char *) program_args.c_str (),
>
> Can we constify fork_inferior's parameters to avoid such
> cast hacks ?

I constified the args as requested, and now most of the casts are gone.
I have a patch here that I'll send later, which constifies more parts of
this code.

>> +		       environ_vector (get_environ ()), linux_ptrace_fun,
>> +		       NULL, NULL, NULL, NULL);
>>  
>>    do_cleanups (restore_personality);
>>  
>> -  linux_add_process (pid, 0);
>> +  linux_update_process (pid);
>>  
>>    ptid = ptid_build (pid, pid, 0);
>> -  new_lwp = add_lwp (ptid);
>> +  new_lwp = update_thread_lwp (ptid);
>>    new_lwp->must_set_ptrace_flags = 1;
>>  
>> +  post_fork_inferior (pid, program_argv);
>> +
>>    return pid;
>>  }
>>  
>
>> +/* See common/common-inferior.h.  */
>> +
>> +char *
>> +get_exec_wrapper (void)
>>  {
>> -  char **new_argv = argv;
>> +  static std::string ret;
>> +  static int initialized_p = 0;
>
> bool.

Fixed.

>> +
>> +  if (wrapper_argv.empty ())
>> +    return NULL;
>>  
>> -  if (wrapper_argv != NULL)
>> +  if (!initialized_p)
>>      {
>> -      int i, count = 1;
>> -
>> -      for (i = 0; wrapper_argv[i] != NULL; i++)
>> -	count++;
>> -      for (i = 0; argv[i] != NULL; i++)
>> -	count++;
>> -      new_argv = XALLOCAVEC (char *, count);
>> -      count = 0;
>> -      for (i = 0; wrapper_argv[i] != NULL; i++)
>> -	new_argv[count++] = wrapper_argv[i];
>> -      for (i = 0; argv[i] != NULL; i++)
>> -	new_argv[count++] = argv[i];
>> -      new_argv[count] = NULL;
>> +      for (std::vector<char *>::iterator i = wrapper_argv.begin ();
>> +	   i != wrapper_argv.end ();
>> +	   ++i)
>> +	ret += *i + std::string (" ");
>
>       for (auto arg : wrapper_argv)
> 	ret += arg + " ";

Fixed.

>> +
>> +      /* Erasing the last whitespace.  */
>
> "Erase".

Fixed.

>> +      ret.erase (ret.end () - 1);
>> +
>> +      initialized_p = 1;
>>      }
>>  
>> +  return (char *) ret.c_str ();
>
> Can the function return const instead?

Fixed.

>> +}
>> +
>> +/* See common/common-inferior.h.  */
>> +
>> +char *
>> +get_exec_file (int err)
>> +{
>> +  if (err && program_argv.empty ())
>> +    error (_("Could not get the exec file."));
>
> I think this error message could be improved.  If I see
> that on the gdbserver terminal, I'll have no clue what
> it means.  "get from where??"  How about just saying
> the same thing gdb says, sans the "Use the ..." part?

Fixed.

>> +  return program_argv[0];
>> +}
>> +
>> +/* See server.h.  */
>> +
>> +struct gdb_environ *
>> +get_environ (void)
>> +{
>> +  return our_environ;
>> +}
>> +
>> +/* See server.h.  */
>> +
>> +void
>> +pre_fork_inferior (std::vector<char *> &argv)
>> +{
>>    if (debug_threads)
>>      {
>> -      int i;
>> -      for (i = 0; new_argv[i]; ++i)
>> -	debug_printf ("new_argv[%d] = \"%s\"\n", i, new_argv[i]);
>> +      int idx = 0;
>> +
>> +      for (char *&i : argv)
>
> Reference to pointer is pointless, it's like pointer to
> pointer.  This should be fine:
>
>       for (char *i : argv)
>
> I'd suggest naming the variable something else, since
> it's not really an iterator.  Like e.g.,:
>
>       for (char *str : argv)
>       for (char *arg : argv)

Thanks for the tip.

>> +	{
>> +	  debug_printf ("new_argv[%d] = \"%s\"\n", idx, i);
>> +	  ++idx;
>
> OTOH, if you're needing/maintaining an index anyway,
> this ends up clearer IMO:
>
>       for (int idx = 0; idx < argv.size(); idx++)
>         debug_printf ("new_argv[%d] = \"%s\"\n", idx, argv[idx]);

Fixed.

>> +	}
>>        debug_flush ();
>>      }
>>  
>
>
>
>>  handle_v_run (char *own_buf)
>>  {
>> -  char *p, *next_p, **new_argv;
>> -  int i, new_argc;
>> +  char *p, *next_p;
>> +  std::vector<char *> new_argv;
>> +  int new_argc;
>> +  int i;
>>  
>>    new_argc = 0;
>>    for (p = own_buf + strlen ("vRun;"); p && *p; p = strchr (p, ';'))
>> @@ -2870,62 +2893,91 @@ handle_v_run (char *own_buf)
>>        new_argc++;
>>      }
>>  
>> -  new_argv = (char **) calloc (new_argc + 2, sizeof (char *));
>> -  if (new_argv == NULL)
>> -    {
>> -      write_enn (own_buf);
>> -      return 0;
>> -    }
>> -
>> -  i = 0;
>> -  for (p = own_buf + strlen ("vRun;"); *p; p = next_p)
>> +  for (i = 0, p = own_buf + strlen ("vRun;"); *p; p = next_p, ++i)
>>      {
>>        next_p = strchr (p, ';');
>>        if (next_p == NULL)
>>  	next_p = p + strlen (p);
>>  
>> -      if (i == 0 && p == next_p)
>> -	new_argv[i] = NULL;
>> +      if (p == next_p)
>> +	new_argv.push_back ("''");
>>        else
>>  	{
>>  	  /* FIXME: Fail request if out of memory instead of dying.  */
>> -	  new_argv[i] = (char *) xmalloc (1 + (next_p - p) / 2);
>> -	  hex2bin (p, (gdb_byte *) new_argv[i], (next_p - p) / 2);
>> -	  new_argv[i][(next_p - p) / 2] = '\0';
>> +	  size_t len = 1 + (next_p - p) / 2;
>> +	  char *s = (char *) xmalloc (len);
>> +	  char *ss = (char *) xmalloc (len * 2);
>> +	  char *tmp_s, *tmp_ss;
>> +	  int need_quote;
>> +
>> +	  hex2bin (p, (gdb_byte *) s, (next_p - p) / 2);
>> +	  s[(next_p - p) / 2] = '\0';
>> +
>> +	  tmp_s = s;
>> +	  tmp_ss = ss;
>> +	  need_quote = 0;
>> +	  while (*tmp_s != '\0')
>> +	    {
>> +	      switch (*tmp_s)
>> +		{
>> +		case '\n':
>> +		  *tmp_ss = '\'';
>> +		  ++tmp_ss;
>> +		  need_quote = 1;
>> +		  break;
>> +
>> +		case '\'':
>> +		  *tmp_ss = '\\';
>> +		  ++tmp_ss;
>> +		  break;
>> +
>> +		default:
>> +		  break;
>> +		}
>> +
>> +	      *tmp_ss = *tmp_s;
>> +	      ++tmp_ss;
>> +	      ++tmp_s;
>> +	    }
>> +
>> +	  if (need_quote)
>> +	    *tmp_ss++ = '\'';
>
> Hmm, is this quoting stuff being moved from somewhere,
> or it is new?

This is new, even though GDB has a lot of places that do the same
thing...

>> +
>> +	  *tmp_ss = '\0';
>> +	  new_argv.push_back (ss);
>> +	  xfree (s);
>>  	}
>>  
>>        if (*next_p)
>>  	next_p++;
>> -      i++;
>>      }
>> -  new_argv[i] = NULL;
>
>>  
>> +  /* Gather information about the environment.  */
>> +  our_environ = make_environ ();
>> +  init_environ (our_environ);
>> +
>>    initialize_async_io ();
>>    initialize_low ();
>> +  /* This is called when initializing inflow on GDB.  */
>
> What is this comment supposed be to telling?

Nothing; this was a leftover from another patch.  Removed.

>> +  have_job_control ();
>>    initialize_event_loop ();
>>    if (target_supports_tracepoints ())
>>      initialize_tracepoint ();
>> @@ -3695,13 +3758,11 @@ captured_main (int argc, char *argv[])
>>        int i, n;
>>  
>>        n = argc - (next_arg - argv);
>> -      program_argv = XNEWVEC (char *, n + 1);
>>        for (i = 0; i < n; i++)
>> -	program_argv[i] = xstrdup (next_arg[i]);
>> -      program_argv[i] = NULL;
>> +	program_argv.push_back (xstrdup (next_arg[i]));
>>  
>
>> -#define create_inferior(program, args) \
>> -  (*the_target->create_inferior) (program, args)
>> +#define create_inferior(program) \
>> +  (*the_target->create_inferior) (program)
>
> program_argv, I suppose.

Fixed.

>> +/* See common/common-utils.h.  */
>> +
>> +std::string
>> +stringify_argv (std::vector<char *> &argv)
>> +{
>> +  std::string ret ("");
>
>    std::string ret;

Fixed.

>> +
>> +  for (std::vector<char *>::iterator i = argv.begin () + 1;
>> +       i != argv.end ();
>> +       ++i)
>> +    ret += *i + std::string (" ");
>
> Should we remove the last empty space?

Yep, always a good practice.  Fixed.

>> +
>> +  return ret;
>> +}
>> diff --git a/gdb/gdbserver/utils.h b/gdb/gdbserver/utils.h
>> index b4ded31..a30d99a 100644
>> --- a/gdb/gdbserver/utils.h
>> +++ b/gdb/gdbserver/utils.h
>> @@ -19,7 +19,17 @@
>>  #ifndef UTILS_H
>>  #define UTILS_H
>>  
>> +#include <vector>
>> +
>>  char *paddress (CORE_ADDR addr);
>>  char *pfildes (gdb_fildes_t fd);
>>  
>> +/* Works like FREEARGV, but with std::vector.  */
>> +extern void free_vector_argv (std::vector<char *> &v);
>
> At some point, freeargv is eliminated, and this comment will
> stay behind.  Can we expand it a bit?

Expanded.

>> +
>> +/* Given a vector of arguments ARGV, return a string equivalent to
>> +   joining all the arguments (starting from ARGV + 1) with a
>> +   whitespace separating them.  */
>> +extern std::string stringify_argv (std::vector<char *> &argv);
>
> const ?
>
> This "starting from ARGV + 1" contract was surprising to me.
> I only noticed this when I saw the implementation.
>
> Does it really make sense to put the program name inside the
> vector instead of on a separate argument if we're always going
> to treat argv[0] differently?
>
> I.e., would:
>
> - ...create_inferior (char *program, char **program_args)
> + ...create_inferior (char *program, std::vector<char *> &program_args)
>
> make more sense?

Right.  I reverted this change, and now we're passing *program as the
first argument again.

>>  static int
>> -win32_create_inferior (char *program, char **program_args)
>> +win32_create_inferior (std::vector<char *> &program_argv)
>>  {
>>  #ifndef USE_WIN32API
>>    char real_path[PATH_MAX];
>> @@ -627,6 +625,9 @@ win32_create_inferior (char *program, char **program_args)
>>    int argc;
>>    PROCESS_INFORMATION pi;
>>    DWORD err;
>> +  char *program = program_argv[0];
>> +  std::string program_args = stringify_argv (program_argv);
>> +  char *args = (char *) program_args.c_str ();
>>  
>>    /* win32_wait needs to know we're not attaching.  */
>>    attaching = 0;
>> @@ -636,6 +637,8 @@ win32_create_inferior (char *program, char **program_args)
>>  
>>    flags = DEBUG_PROCESS | DEBUG_ONLY_THIS_PROCESS;
>>  
>> +  pre_fork_inferior (program, argv);
>
> Hmm, fork on Windows?

Ops.  Removed.

>> +
>>  #ifndef USE_WIN32API
>>    orig_path = NULL;
>>    path_ptr = getenv ("PATH");
>> @@ -652,18 +655,6 @@ win32_create_inferior (char *program, char **program_args)
>>    program = real_path;
>>  #endif
>>  
>> -  argslen = 1;
>> -  for (argc = 1; program_args[argc]; argc++)
>> -    argslen += strlen (program_args[argc]) + 1;
>> -  args = (char *) alloca (argslen);
>> -  args[0] = '\0';
>> -  for (argc = 1; program_args[argc]; argc++)
>> -    {
>> -      /* FIXME: Can we do better about quoting?  How does Cygwin
>> -	 handle this?  */
>> -      strcat (args, " ");
>> -      strcat (args, program_args[argc]);
>> -    }
>>    OUTMSG2 (("Command line is \"%s\"\n", args));
>>  
>>  #ifdef CREATE_NEW_PROCESS_GROUP
>> @@ -704,6 +695,8 @@ win32_create_inferior (char *program, char **program_args)
>>  
>>    do_initial_child_stuff (pi.hProcess, pi.dwProcessId, 0);
>>  
>> +  post_fork_inferior (current_process_id, program_argv);
>> +
>>    return current_process_id;
>>  }
>>  
>
>
>> @@ -39,8 +39,14 @@ set spawn_id [remote_spawn target "$gdbserver stdio non-existing-program"]
>>  set msg "gdbserver exits cleanly"
>>  set saw_exiting 0
>>  expect {
>> -    # This is what we get on ptrace-based targets.
>> -    -re "stdin/stdout redirected.*No program to debug\r\nExiting\r\n$" {
>> +    # This is what we get on ptrace-based targets with
>> +    # startup-with-shell disabled.
>> + -re "stdin/stdout redirected.*gdbserver: Cannot exec
>> non-existing-program\r\ngdbserver: Error: No such file or
>> directory\r\n\r\nDuring startup program exited with code
>> 127\.\r\nExiting\r\n$" {
>> +	set saw_exiting 1
>> +	exp_continue
>
> Shouldn't this be a part of the next patch?

Not really.  I put this here because without it a regreession is
introduced, and I wanted each patch to be regression-free.

>> +    }
>> +    # Likewise, but with startup-with-shell enabled.
>> +    -re "stdin/stdout redirected.*exec: non-existing-program: not found\r\nDuring startup program exited with code 127\.\r\nExiting\r\n$" {
>>  	set saw_exiting 1
>>  	exp_continue
>>      }

I'll send the next version soon.

Thanks,

-- 
Sergio
GPG key ID: 237A 54B1 0287 28BF 00EF  31F4 D0EB 7628 65FC 5E36
Please send encrypted e-mail if possible
http://sergiodj.net/


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]