On Sun 01 Apr 2001, Geoff Wing wrote:
> Paul Slootman <[EMAIL PROTECTED]> typed:
> : On Sat 30 Sep 2000, Geoff Wing wrote:
> : It appears that "test -e" isn't standard; "-c" might be a better choice
> : (I'm guessing /dev/ptym/clone is a character special device).
> : After fixing this, it reports
> : pty/tty type: GLIBC
> : which makes me wonder a bit, I don't think solaris uses glibc...
Now it reports
pty/tty type: STREAMS
which looks a lot better!
> Second patch makes sure /dev/tty is reassociated with the tty (as
> a controlling tty) when certain OS-specific ioctl()s aren't available.
> diff -u -r1.69 init.c
> --- src/init.c 2001/03/23 06:15:37 1.69
> +++ src/init.c 2001/04/01 08:12:27
> @@ -1137,6 +1137,12 @@
> ioctl(r->tty_fd, TIOCSCTTY, 0);
> # elif defined(TIOCSETCTTY)
> ioctl(r->tty_fd, TIOCSETCTTY, 0);
> +# else
> + fd = open(r->h->ttydev, O_RDWR);
Make this
int fd = open(r->h->ttydev, O_RDWR);
^^^
otherwise it complains about fd not being declared...
It compiles with this, but still doesn't work. It now dies without any
error; DEBUG_TTY says:
ontrol_tty(): pid: 8783, tty fd: 6, dev: /dev/pts/15
rxvt_control_tty(): /dev/tty has controlling tty? no (good)
rxvt_control_tty(): tty opened OK
rxvt_control_tty(): do we have controlling tty now: yes (good)
rxvt_control_tty(): tcgetpgrp(): 8783 getpgrp(): 8783
and that's the last I see.
Running with -e sh -c '/bin/stty -a; sleep 100' I see a window opened,
and the following output:
stty: : Invalid argument
(the cursor is below the end of the text, i.e. -opost :-)
truss says:
8980: ioctl(0, STGET, 0x0002643C) Err#22 EINVAL
8980: ioctl(0, TCGETS, 0x00026460) Err#22 EINVAL
8980: ioctl(0, TCGETA, 0x0002644C) Err#22 EINVAL
8980: write(2, " s t t y : ", 6) = 6
8972: poll(0xEFFFCF30, 2, -1) = 1
8980: write(2, " : ", 2) = 2
8972: read(5, " s t t y : : ", 1024) = 8
8972: read(5, " I n v a l i d a r g u".., 1016) = 16
8980: write(2, " I n v a l i d a r g u".., 16) = 16
Aha! The streams modules aren't being pushed. The following patches
are needed (includes the fd thing I already mentioned):
--- rxvt.cvs/src/init.c Mon Apr 2 13:19:05 2001
+++ rxvt/src/init.c Mon Apr 2 13:23:26 2001
@@ -1143,7 +1143,7 @@
ioctl(STDIN_FILENO, TIOCCONS, &on);
#elif defined (SRIOCSREDIR)
- fd = open(CONSOLE, O_WRONLY, 0);
+ int fd = open(CONSOLE, O_WRONLY, 0);
if (fd >= 0) {
if (ioctl(fd, SRIOCSREDIR, NULL) < 0)
close(fd);
--- rxvt.cs/src/init.h Fri Mar 23 07:15:37 2001
+++ rxvt/src/init.h Mon Apr 2 13:45:40 2001
@@ -16,7 +16,6 @@
#if defined(PTYS_ARE_PTMX) && !defined(__CYGWIN32__)
# include <sys/resource.h> /* for struct rlimit */
-# include <sys/stropts.h> /* for I_PUSH */
# define _NEW_TTY_CTRL /* to get proper defines in <termios.h> */
#endif
--- rxvt.cvs/src/ptytty.c Mon Apr 2 09:32:35 2001
+++ rxvt/src/ptytty.c Mon Apr 2 13:45:55 2001
@@ -23,6 +23,10 @@
#include "../config.h" /* NECESSARY */
+#if defined(PTYS_ARE_PTMX) && !defined(__CYGWIN32__)
+# include <sys/stropts.h> /* for I_PUSH */
+#endif
+
/*
* Try to be self-contained except for the above autoconfig'd defines
*/
This seems to make it work.
BTW, my unscientific timings of the 2.6.2 version and the current CVS:
time locate / | wc
81819 81819 5323527
real 0m0.259s
user 0m0.400s
sys 0m0.060s
I.e. 5MB over 81000 lines, locate itself barely using CPU.
[2.6.2] time /usr/local/bin/rxvt -e locate /
real 0m24.653s
user 0m2.650s
sys 0m2.020s
[CVS] time src/rxve -e locate /
real 0m7.467s
user 0m2.890s
sys 0m1.950s
This is on a remote display (Xoftware from Chameleon on win95), 500
scrollback lines, 24x80, default font size).
Good work!
Paul Slootman
--
home: [EMAIL PROTECTED] http://www.wurtel.demon.nl/
work: [EMAIL PROTECTED] http://www.murphy.nl/
debian: [EMAIL PROTECTED] http://www.debian.org/
isdn4linux: [EMAIL PROTECTED] http://www.isdn4linux.org/
