658     /* In posix_spawn mode, require the child process to signal aliveness

   659      * right after it comes up. This is because there are implementations of

   660      * posix_spawn() which do not report failed exec()s back to the caller

   661      * (e.g. glibc, see JDK-8223777). In those cases, the fork() will have

   662      * worked and successfully started the child process, but the exec() will

   663      * have failed. There is no way for us to distinguish this from a target

   664      * binary just exiting right after start.

   665      *

   666      * Note that we could do this additional handshake in all modes but for

   667      * prudence only do it when it is needed (in posix_spawn mode). */

   668     c->sendAlivePing = (mode == MODE_POSIX_SPAWN) ? 1 : 0;

   669 

   689     /* If we expect the child to ping aliveness, wait for it. */

   690     if (c->sendAlivePing) {

   691         switch(readFully(fail[0], &errnum, sizeof(errnum))) {

   692         case 0: /* First exec failed; */

   693             waitpid(resultPid, NULL, 0);

   694             throwIOException(env, 0, "Failed to exec spawn helper.");

   695             goto Catch;

   696         case sizeof(errnum):

   697             assert(errnum == CHILD_IS_ALIVE);

   698             if (errnum != CHILD_IS_ALIVE) {

   699                 /* Should never happen since the first thing the spawn

   700                  * helper should do is to send an alive ping to the parent,

   701                  * before doing any subsequent work. */

   702                 throwIOException(env, 0, "Bad code from spawn helper "

   703                                          "(Failed to exec spawn helper.");

   704                 goto Catch;

   705             }

   706             break;

   707         default:

   708             throwIOException(env, errno, "Read failed");

   709             goto Catch;

   710         }

   711     }

   712