> With your diff, the expansion will be done only once, on the initial
> assignment of PS1. I do not see how you'd get the current working dir
> in a prompt with your diff. And what's more, I do not see how the
> current behaviour is non-posix, as you claim.

Yes, you are right and I am wrong. Sorry about that. I was confused
why it is saying

PS1="$x$(print \\r)$x$(tput so)$x\$PWD$x$(tput se)$x> "

and not directly

PS1='$x$(print \\r)$x$(tput so)$x$PWD$x$(tput se)$x> '

but in the end both has the same effect. And after reading that posix
quote again, you are also right, it *shall* be parameter expanded,
which gets lost with my last diff.

How about this diff. I've updated the regression test a bit which
you can fetch here:

http://www.nazgul.ch/dev/ksh_regress.tar.gz

Regards,
Marcus

--
Marcus Glocker, marcus@nazgul.ch, http://www.nazgul.ch -----------------
diff -urN -x CVS src/bin/ksh.orig/lex.c src/bin/ksh/lex.c
--- src/bin/ksh.orig/lex.c Mon Apr 10 16:38:59 2006
+++ src/bin/ksh/lex.c Thu May 25 15:53:14 2006
@@ -67,6 +67,7 @@

static int backslash_skip;
static int ignore_backslash_newline;
+static int ps1_flag;

/* optimized getsc_bn() */
#define getsc() (*source->str != '\0' && *source->str != '\\' \
@@ -288,6 +289,18 @@
switch (c) {
case '\\':
c = getsc();
+ /*
+ * For currently active PS1 string don't do
+ * escape character substitution anymore
+ * for those characters which are used in
+ * dopprompt()
+ */
+ if (ps1_flag) {
+ if (c == '$') {
+ *wp++ = '\\', *wp++ = c;
+ break;
+ }
+ }
switch (c) {
case '\\':
case '$': case '`':
@@ -1141,9 +1154,12 @@
* unwinding its stack through this code as it
* exits.
*/
- } else
+ } else {
+ ps1_flag = 1;
prompt = str_save(substitute(ps1, 0),
saved_atemp);
+ ps1_flag = 0;
+ }
quitenv(NULL);
break;
case PS2: /* command continuation */