From: Simon Strandgaard <0bz63fz3m1qt3001@...>
Date: 2003-05-13T18:30:37+09:00
Subject: Re: RCR for child execution

On Tue, 13 May 2003 18:43:28 +0900, Yukihiro Matsumoto wrote:

> Hi,
> 
> In message "Re: RCR for child execution"
>     on 03/05/13, "Simon Strandgaard" <0bz63fz3m1qt3001@sneakemail.com> writes:
> 
> |Still rb_warn and friends outputs to stderr. 
> |I am searching for the nice solution :-)
> 
> Now it does.
> 
> In message "Re: RCR for child execution"
>     on 03/05/13, "Simon Strandgaard" <0bz63fz3m1qt3001@sneakemail.com> writes:
> 
> |void system(command) {
> |    pid=fork();
> |    if(pid != 0) {
> |         stdout = rb_stdout;  /* what im asking for */
> |         stderr = rb_stderr;  /* what im asking for */
> |         stdin = rb_stdin;    /* what im asking for */
> |         execv(command);
> |    }
> |}
> |
> |OK.. again assignment to stdout is illegal. freopen should be used instead.
> |Admitted.. I havn't yet tried reopen'ing stdout/stderr :-)
> 
> You misunderstand here.  In current Ruby, assignment to $stdin
> etc. does reopen *at the time of assignment*.

Yes I know rb_stdout is a hooked variable.. this is just very simplified
pseudo code. 
Actually I have a bug in this.. stdout is suppose to inherit rb_stdout
from the parent process. So here is my example I posted earlier:

void system(command) {
    backup_stdout = stdout;
    stdout = rb_stdout;
    pid = fork();
    if(pid != 0)
        execv(command);
    stdout = backup_stdout;
}

Again its pseudo code :-)
This is only parsing rb_stdout from parent to child... of cause also
rb_stderr and rb_stdin should be parsed to the child too.

--
Simon Strandgaard