Am Freitag, 2. April 2004 19:45 schrieb Juergen Spitzmueller:
> errno is ECHILD (no child processes), so your guess is probably right.
> OTOH man popen says: "If pclose() cannot obtain the child status, errno is
set
> to ECHILD", which sounds a bit vague.
Yes, it seems that it does not tell all information we need. And the sun
manpage that you cited in your other mail tells something similar.
> Is the attached safe? It sets pret to 0 when errno is ECHILD and fixes the
> bug. It also prints out an error message both when popen does not succeed
and
> when pclose gets an error (other than ECHILD).
As I understand it, it is not safe to assume that the child process was
executed successfully if errno is ECHILD. The siginfo struct that can be
obtained with sigwaitinfo() has probably the needed information. On linux,
this struct and the following enum is defined in
/usr/include/bits/siginfo.h:
/* `si_code' values for SIGCHLD signal. */
enum
{
CLD_EXITED = 1, /* Child has exited. */
# define CLD_EXITED CLD_EXITED
CLD_KILLED, /* Child was killed. */
# define CLD_KILLED CLD_KILLED
CLD_DUMPED, /* Child terminated abnormally. */
# define CLD_DUMPED CLD_DUMPED
CLD_TRAPPED, /* Traced child has trapped. */
# define CLD_TRAPPED CLD_TRAPPED
CLD_STOPPED, /* Child has stopped. */
# define CLD_STOPPED CLD_STOPPED
CLD_CONTINUED /* Stopped child has continued. */
# define CLD_CONTINUED CLD_CONTINUED
};
If we had the value of si_code, we could ignore the error for CLD_EXITED.
But I don't know if/how/when one could call sigwaitinfo() to obtain the
desired information.
Georg
