Unix98 Pty Support for XEmacs
24 years, 12 months
Andreas Jaeger
Andreas Jaeger
The following patch against 21.1.7 adds support for Unix98 Ptys
(/dev/ptmx) as supported by Linux 2.2 kernels and glibc 2.1.
IMO the patch has one problem: It checks directly for __GLIBC__ and
only works for __GLIBC__. Replacing getpt with open ("/dev/ptmx",..)
and ptsname_r with ptsname would get this working on otherplatforms
(getpt and ptsname_r are glibc extensions - but getpt is quite useful
since it even works if /dev/ptmx doesn't work;-).
I guess, I should rewrite this and use some configure tests. Since
opening /dev/ptmx might involve some system specific handling, I would
check for now for grantpt, unlockpt and ptsname - and if all exists, define
USE_UNIX98_PTYS - and check additionally for getpt - and use
USE_UNIX98_PTYS && HAVE_GETPT instead of the check for glibc.
What do you think?
Andreas
--- process-unix.c
+++ process-unix.c Wed Dec 1 17:00:02 1999
＠＠ -204,6 +204,21 ＠＠
static int
allocate_pty (void)
{
+ int fd;
+#if __GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 1)
+ if ((fd = getpt ()) < 0)
+ return -1;
+
+ if (grantpt (fd) < 0 ||
+ unlockpt (fd) < 0 ||
+ ptsname_r (fd, pty_name, sizeof pty_name))
+ {
+ close (fd);
+ return -1;
+ }
+ setup_pty (fd);
+ return fd;
+#else /* !glibc */
#ifndef PTY_OPEN
struct stat stb;
＠＠ -215,7 +230,6 ＠＠
int failed_count = 0;
#endif
int i;
- int fd;
int c;
#ifdef PTY_ITERATION
＠＠ -283,6 +297,7 ＠＠
}
}
return -1;
+#endif /* !glibc */
}
#endif /* HAVE_PTYS */
--
Andreas Jaeger
SuSE Labs aj(a)suse.de
private aj(a)arthur.rhein-neckar.de
- 1 participants
- 0 comments
-b20 and -b22 are leaking tab widgets
24 years, 12 months
Karl Kleinpaste
Karl Kleinpaste
Having reported the problems where 21.2-b20 and -b22 both exit(1) out
of Xt due to invalid widget sizes, and having noticed that "xwininfo
-tree" reports seemingly absurd child window counts in those cases, I
watched a bit more closely today.
-b20 and -b22 are leaking widgets when running Gnus.
Every time I enter a new group, "xwininfo -tree" shows a raised-by-one
count of children of the first window. Every time I show the first
article in the freshly-entered group, that same count increases by 3
more. My usual display in Gnus is to show Group, Summary, Article,
and Tree. The count then stays constant as long as I continue reading
articles in a given group. Leave the group, no change. Enter the
next group, and the leakage begins again, with a raised-by-one count.
Read the first article, and... So every time I move steadily through
a group, the child count of the main window goes up by a total of 4.
If I enter a group but then leave it without reading any articles,
total increase is just 1, of course.
Turning off tabs stops the leakage dead in its tracks. But the count
doesn't go down, either -- items leaked are never recovered.
Linux RH6.0, XFree 3.3.5, "./configure --with-pop --with-mule".
BTW, I put Jan's 1-line lwlib-Xaw.c update to use, and haven't seen
the exit(1) since; but as of a moment ago, I've turned off tabs
entirely to avoid the leakage, so now I'm not really learning whether
Jan's fix is genuinely helpful.
--karl
- 1 participants
- 0 comments
up to date sumo package?
24 years, 12 months
Cotton Seed
Cotton Seed
The latest xemacs-sumo tarball in ftp.xemacs.org:/pub/xemacs/packages
is dated July 29, but it is named xemacs-sumo.tar.gz, not
xemacs-sumo-<date>.tar.gz, as described in README.packages. There is
a xemacs-sumo tarball so named dated July 13. This is about 5 months
old. Is this really up to date? If not, are there plans to build a
up to date xemacs-sumo tarball?
- Cotton
- 2 participants
- 1 comments
Using mktemp for temporary files
24 years, 12 months
Andreas Jaeger
Andreas Jaeger
A couple of scripts create temporary files - but those names are easy
to guess and could lead to security problems. A better solution would
be to use mktemp.
I've got here a patch which does the following:
--- lib-src/rcs2log
+++ lib-src/rcs2log Thu Nov 11 17:41:36 1999
＠＠ -279,6 +279,15 ＠＠
llogout=$TMPDIR/rcs2log$$l
rlogout=$TMPDIR/rcs2log$$r
+rm -f $llogout $rlogout
+if test -e $llogout -o -e $rlogout ; then
+ echo "$0: temporary files $llogout or/and $rlogout exists." 1>&2
+ exit 1
+fi
+if test -n "`type -p mktemp`" ; then
+ llogout="`mktemp ${llogout}.XXXXXX`" || exit 1
+ rlogout="`mktemp ${rlogout}.XXXXXX`" || exit 1
+fi
trap exit 1 2 13 15
trap "rm -f $llogout $rlogout; exit 1" 0
What strikes me is that I don't know whether type -p is available in
/bin/sh.
Besides, does anybody have better ideas to rewrite this. There are a
number of other scripts which needs rewriting. I've some likes to do
this, I can forward the complete patch set I have.
Andreas
--
Andreas Jaeger
SuSE Labs aj(a)suse.de
private aj(a)arthur.rhein-neckar.de
- 4 participants
- 3 comments