/* pae@dim.com (Phil Ernhardt) says this correction to
the definition in usg5-4.h is needed to prevent
all asynchronous subprocesses from exiting right away. */
+
+/* James Youngman <jay@gnu.org> found that on "DYNIX/ptx ARNIE 4.0
+ * V4.4.2 i386", the push of the ttcompat module would fail. It seems
+ * that PTX 4.4.2 doesn't have that module, so if the push fails we
+ * don't kill ourselves. While this version lacks ttcompat, it also
+ * has ptem, but the manual page for ptem indicates that it should
+ * be pushed onto the slave side before the line discipline module.
+ * See also the streampty manual page, if you're curious (and have
+ * a ptx system).
+ */
+
+/* rms: I hope that older versions which do have ttcompat
+ will not get confused by the code to use ptem. */
+
#undef SETUP_SLAVE_PTY
#define SETUP_SLAVE_PTY \
+ if (ioctl (xforkin, I_PUSH, "ptem") == -1 && errno != EINVAL) \
+ fatal ("ioctl I_PUSH ptem", errno); \
if (ioctl (xforkin, I_PUSH, "ldterm") == -1) \
fatal ("ioctl I_PUSH ldterm", errno); \
- if (ioctl (xforkin, I_PUSH, "ttcompat") == -1) \
+ if (ioctl (xforkin, I_PUSH, "ttcompat") == -1 && errno != EINVAL) \
fatal ("ioctl I_PUSH ttcompat", errno);