Help on pipe

Help on pipe

[Ardhan Madras]

Hi,

I'm writing pipe demo:

#include <sys/types.h>
#include <sys/wait.h>
#include <unistd.h>
#include <stdio.h>
#include <string.h>

int main(void)
{
  pid_t pid;
  int fds[2], ret;
  const char *sort = "/usr/bin/sort";
  char *argv[2] = { "sort", NULL };

  ret = pipe(fds);
  if (ret == -1) {
    perror("pipe");
    return -1;
  }
  pid = fork();
  if (pid == -1) {
    perror("fork");
    return -1;
  }

  if (pid == 0) {
    int ret;

    close(fds[1]);
    dup2(fds[0], STDIN_FILENO);

    ret = execve(sort, argv, NULL);
    if (ret == -1) {
      perror("execve");
    }
    close(fds[0]);
    _exit(0);
  }
  else {
    FILE *stream;

    close(fds[0]);
    stream = fdopen(fds[1], "w");
    if (!stream) {
      perror("fdopen");
      return -1;
    }
    fprintf(stream, "this\n");
    fprintf(stream, "means\n");
    fprintf(stream, "war\n");
    fflush(stream);
    waitpid(pid, &ret, 0);
    close(fds[1]);
  }
  return 0;
}

I want to send parent's data to the child as "sort" input, i don't
know why sort doesn't receive the data. Did i made mistake here?

Thanks before.

Re: Help on pipe

[Michał Nazarewicz]

On Fri, 10 Jul 2009 08:22:54 +0200, Ardhan Madras <nightdecoder@gmail.com>  
wrote:

> #include <sys/types.h>
> #include <sys/wait.h>
> #include <unistd.h>
> #include <stdio.h>
> #include <string.h>
>
> int main(void)
> {
>   pid_t pid;
>   int fds[2], ret;
>   const char *sort = "/usr/bin/sort";

"static const char sort[]" would be better IMO.

>   char *argv[2] = { "sort", NULL };

"static const char *argv[]" is definitly better here and does not produce  
warnings.  Your code discards "const" from pointer (a string literal).

>   ret = pipe(fds);
>   if (ret == -1) {
>     perror("pipe");
>     return -1;
>   }
>   pid = fork();
>   if (pid == -1) {
>     perror("fork");
>     return -1;
>   }

Personally I preffer something like:

   switch (fork()) {
   case -1: /* error */  break;
   case  0: /* child */  break;
   default: /* parent */
   }

but suit yourself.

>   if (pid == 0) {
>     int ret;

No need to define this variable here.  You already have such variable.

>
>     close(fds[1]);
>     dup2(fds[0], STDIN_FILENO);
>
>     ret = execve(sort, argv, NULL);
>     if (ret == -1) {
>       perror("execve");
>     }
>     close(fds[0]);
>     _exit(0);

No need to close(), and _exit(0) should be _exit(1) (or -1 if you will)  
since the chiled finished with error.

>   }
>   else {
>     FILE *stream;
>
>     close(fds[0]);
>     stream = fdopen(fds[1], "w");
>     if (!stream) {
>       perror("fdopen");
>       return -1;
>     }
>     fprintf(stream, "this\n");
>     fprintf(stream, "means\n");
>     fprintf(stream, "war\n");
>     fflush(stream);

And here, replace fflush() with fclose().  It seems, data is buffered on  
kernel level, not only libc level and fflush() cannot handle that.

>     waitpid(pid, &ret, 0);
>     close(fds[1]);

No need to close.

>   }
>   return 0;
> }

-- 
Best regards,                                            _     _
  .o. | Liege of Serenly Enlightened Majesty of         o' \,=./ `o
  ..o | Computer Science,  Michał "mina86" Nazarewicz      (o o)
  ooo +-<m.nazarewicz@samsung.com>-<mina86@jabber.org>-ooO--(_)--Ooo--

--
To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Help on pipe

[David Lee]

On Fri, Jul 10, 2009 at 2:22 PM, Ardhan Madras<nightdecoder@gmail.com> wrote:
> Hi,
>
> I'm writing pipe demo:
>
> #include <sys/types.h>
> #include <sys/wait.h>
> #include <unistd.h>
> #include <stdio.h>
> #include <string.h>
>
> int main(void)
> {
>  pid_t pid;
>  int fds[2], ret;
>  const char *sort = "/usr/bin/sort";
>  char *argv[2] = { "sort", NULL };
>
>  ret = pipe(fds);
>  if (ret == -1) {
>    perror("pipe");
>    return -1;
>  }
>  pid = fork();
>  if (pid == -1) {
>    perror("fork");
>    return -1;
>  }
>
>  if (pid == 0) {
>    int ret;
>
>    close(fds[1]);
>    dup2(fds[0], STDIN_FILENO);
>
>    ret = execve(sort, argv, NULL);
>    if (ret == -1) {
>      perror("execve");
>    }
>    close(fds[0]);
>    _exit(0);
>  }
>  else {
>    FILE *stream;
>
>    close(fds[0]);
>    stream = fdopen(fds[1], "w");
>    if (!stream) {
>      perror("fdopen");
>      return -1;
>    }
>    fprintf(stream, "this\n");
>    fprintf(stream, "means\n");
>    fprintf(stream, "war\n");
>    fflush(stream);
>    waitpid(pid, &ret, 0);
>    close(fds[1]);
>  }
>  return 0;
> }
>
> I want to send parent's data to the child as "sort" input, i don't
> know why sort doesn't receive the data. Did i made mistake here?
>
> Thanks before.
> --
> To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

The sort program is waiting for an EOF (^D).

--- sort.c.orig 2009-07-10 15:54:36.000000000 +0800
+++ sort.c      2009-07-10 15:53:35.000000000 +0800
@@ -49,6 +49,7 @@
                fprintf(stream, "means\n");
                fprintf(stream, "war\n");
                fflush(stream);
+              fclose(stream);
                waitpid(pid, &ret, 0);
                close(fds[1]);
        }


-- 
Thanks,
Li Qun
--
To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Help on pipe

[Ardhan Madras]

My demo can run. i use -Wall flag, the "const char *argv[2] = {
"sort", NULL }" as you said warned with:

pipe.c:35: warning: passing argument 2 of 'execve' from incompatible
pointer type.

but it's perfect for  "char * const argv[2] = { "sort", NULL }"

and i define "int ret" for both process, but i undestand forking
mechanism on GNU/Linux.

sorry i don't write the code cleanly.. My point is only to make the
"sort" working, im now understand that "sort" will always wait for
input until it's end (peer closed the input).

Thank you very much.



2009/7/10 Micha³ Nazarewicz <m.nazarewicz@samsung.com>:
> On Fri, 10 Jul 2009 08:22:54 +0200, Ardhan Madras <nightdecoder@gmail.com>
> wrote:
>
>> #include <sys/types.h>
>> #include <sys/wait.h>
>> #include <unistd.h>
>> #include <stdio.h>
>> #include <string.h>
>>
>> int main(void)
>> {
>>  pid_t pid;
>>  int fds[2], ret;
>>  const char *sort = "/usr/bin/sort";
>
> "static const char sort[]" would be better IMO.
>
>>  char *argv[2] = { "sort", NULL };
>
> "static const char *argv[]" is definitly better here and does not produce
> warnings.  Your code discards "const" from pointer (a string literal).
>
>>  ret = pipe(fds);
>>  if (ret == -1) {
>>    perror("pipe");
>>    return -1;
>>  }
>>  pid = fork();
>>  if (pid == -1) {
>>    perror("fork");
>>    return -1;
>>  }
>
> Personally I preffer something like:
>
>  switch (fork()) {
>  case -1: /* error */  break;
>  case  0: /* child */  break;
>  default: /* parent */
>  }
>
> but suit yourself.
>
>>  if (pid == 0) {
>>    int ret;
>
> No need to define this variable here.  You already have such variable.
>
>>
>>    close(fds[1]);
>>    dup2(fds[0], STDIN_FILENO);
>>
>>    ret = execve(sort, argv, NULL);
>>    if (ret == -1) {
>>      perror("execve");
>>    }
>>    close(fds[0]);
>>    _exit(0);
>
> No need to close(), and _exit(0) should be _exit(1) (or -1 if you will)
> since the chiled finished with error.
>
>>  }
>>  else {
>>    FILE *stream;
>>
>>    close(fds[0]);
>>    stream = fdopen(fds[1], "w");
>>    if (!stream) {
>>      perror("fdopen");
>>      return -1;
>>    }
>>    fprintf(stream, "this\n");
>>    fprintf(stream, "means\n");
>>    fprintf(stream, "war\n");
>>    fflush(stream);
>
> And here, replace fflush() with fclose().  It seems, data is buffered on
> kernel level, not only libc level and fflush() cannot handle that.
>
>>    waitpid(pid, &ret, 0);
>>    close(fds[1]);
>
> No need to close.
>
>>  }
>>  return 0;
>> }
>
> --
> Best regards,                                            _     _
>  .o. | Liege of Serenly Enlightened Majesty of         o' \,=./ `o
>  ..o | Computer Science,  Micha³ "mina86" Nazarewicz      (o o)
>  ooo +-<m.nazarewicz@samsung.com>-<mina86@jabber.org>-ooO--(_)--Ooo--
>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Help on pipe

[Glynn Clements]

Micha³ Nazarewicz wrote:

> >     waitpid(pid, &ret, 0);
> >     close(fds[1]);
> 
> No need to close.

There definitely is a need to close, but it should be fclose(stream)
*before* the waitpid(). As David points out, sort won't terminate
until it has finished reading its input.

David Lee wrote:

> The sort program is waiting for an EOF (^D).

Ctrl-D isn't EOF unless you're reading from a terminal. The TTY driver
"sends EOF" (i.e. causes read() to return a zero count) when you press
Ctrl-D (by default) if the driver is in canonical mode. Ctrl-D doesn't
have any significance in other contexts.

-- 
Glynn Clements <glynn@gclements.plus.com>
--
To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Help on pipe

[David Lee]

2009/7/11 Glynn Clements <glynn@gclements.plus.com>:
>
> Micha³ Nazarewicz wrote:
>
>> >     waitpid(pid, &ret, 0);
>> >     close(fds[1]);
>>
>> No need to close.
>
> There definitely is a need to close, but it should be fclose(stream)
> *before* the waitpid(). As David points out, sort won't terminate
> until it has finished reading its input.
>
> David Lee wrote:
>
>> The sort program is waiting for an EOF (^D).
>
> Ctrl-D isn't EOF unless you're reading from a terminal. The TTY driver
> "sends EOF" (i.e. causes read() to return a zero count) when you press
> Ctrl-D (by default) if the driver is in canonical mode. Ctrl-D doesn't
> have any significance in other contexts.
>
Definitely!  Thanks for pointing out. :-)
And BTW, `strace' is a handy tool for diagnosing this kind of problems.

> --
> Glynn Clements <glynn@gclements.plus.com>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

-- 
Thanks,
Li Qun
--
To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Help on pipe

[Michał Nazarewicz]

>>>     waitpid(pid, &ret, 0);
>>>     close(fds[1]);

> Michał Nazarewicz wrote:
>> No need to close.

On Fri, 10 Jul 2009 19:40:38 +0200, Glynn Clements  
<glynn@gclements.plus.com> wrote:
> There definitely is a need to close, but it should be fclose(stream)
> *before* the waitpid(). As David points out, sort won't terminate
> until it has finished reading its input.

Yes, yes... David already pointed out the problem is with EOF not  
buffering.

-- 
Best regards,                                            _     _
  .o. | Liege of Serenly Enlightened Majesty of         o' \,=./ `o
  ..o | Computer Science,  Michał "mina86" Nazarewicz      (o o)
  ooo +-<m.nazarewicz@samsung.com>-<mina86@jabber.org>-ooO--(_)--Ooo--

--
To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Help on pipe

[Ardhan Madras]

My demo can run. i use -Wall flag, the "const char *argv[2] = {
"sort", NULL }" as you said warned with:

pipe.c:35: warning: passing argument 2 of 'execve' from incompatible
pointer type.

but it's perfect for  "char * const argv[2] = { "sort", NULL }"

and i define "int ret" for both process, but i undestand forking
mechanism on GNU/Linux.

sorry i don't write the code cleanly.. My point is only to make the
"sort" working, im now understand that "sort" will always wait for
input until it's end (peer closed the input).

Thank you very much.



2009/7/10 Michał Nazarewicz <m.nazarewicz@samsung.com>:
> On Fri, 10 Jul 2009 08:22:54 +0200, Ardhan Madras <nightdecoder@gmail.com>
> wrote:
>
>> #include <sys/types.h>
>> #include <sys/wait.h>
>> #include <unistd.h>
>> #include <stdio.h>
>> #include <string.h>
>>
>> int main(void)
>> {
>>  pid_t pid;
>>  int fds[2], ret;
>>  const char *sort = "/usr/bin/sort";
>
> "static const char sort[]" would be better IMO.
>
>>  char *argv[2] = { "sort", NULL };
>
> "static const char *argv[]" is definitly better here and does not produce
> warnings.  Your code discards "const" from pointer (a string literal).
>
>>  ret = pipe(fds);
>>  if (ret == -1) {
>>    perror("pipe");
>>    return -1;
>>  }
>>  pid = fork();
>>  if (pid == -1) {
>>    perror("fork");
>>    return -1;
>>  }
>
> Personally I preffer something like:
>
>  switch (fork()) {
>  case -1: /* error */  break;
>  case  0: /* child */  break;
>  default: /* parent */
>  }
>
> but suit yourself.
>
>>  if (pid == 0) {
>>    int ret;
>
> No need to define this variable here.  You already have such variable.
>
>>
>>    close(fds[1]);
>>    dup2(fds[0], STDIN_FILENO);
>>
>>    ret = execve(sort, argv, NULL);
>>    if (ret == -1) {
>>      perror("execve");
>>    }
>>    close(fds[0]);
>>    _exit(0);
>
> No need to close(), and _exit(0) should be _exit(1) (or -1 if you will)
> since the chiled finished with error.
>
>>  }
>>  else {
>>    FILE *stream;
>>
>>    close(fds[0]);
>>    stream = fdopen(fds[1], "w");
>>    if (!stream) {
>>      perror("fdopen");
>>      return -1;
>>    }
>>    fprintf(stream, "this\n");
>>    fprintf(stream, "means\n");
>>    fprintf(stream, "war\n");
>>    fflush(stream);
>
> And here, replace fflush() with fclose().  It seems, data is buffered on
> kernel level, not only libc level and fflush() cannot handle that.
>
>>    waitpid(pid, &ret, 0);
>>    close(fds[1]);
>
> No need to close.
>
>>  }
>>  return 0;
>> }
>
> --
> Best regards,                                            _     _
>  .o. | Liege of Serenly Enlightened Majesty of         o' \,=./ `o
>  ..o | Computer Science,  Michał "mina86" Nazarewicz      (o o)
>  ooo +-<m.nazarewicz@samsung.com>-<mina86@jabber.org>-ooO--(_)--Ooo--
>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html




_____________________________________________________________
Listen to KNAC, Hit the Home page and Tune In Live! ---> http://www.knac.com