Hi!
On Tue, Oct 25, 2005 at 04:40:53AM -0700, Steve Langasek wrote:
> > > Sorry, what problem are you trying to solve?
>
> > Excuse me for the confusion. I talk here about another
> > issue, similar to what we have in bug #333138.
>
> > Now I tend to agree with Robert that this is a bug,
> > because fact of stdin being or not being tty should not
> > affect buffering/flushing of stdout/stderr. At least
> > this behaviour is rather strange and unexpected to
> > majority of *nix users, admins and programmers.
>
> It's not unusual for password-related tools to insist on a tty; though
> granted, misc_conv allows non-tty conversations, so this is still a bug,
> just not one I'd consider very serious. Anyway, if you can send me a patch,
> I'll see about getting a fix pushed upstream.
OK, and anyway I think the change is quite trivial and
safe. Thus I attach it here. It's against upstream's
source (Linux-PAM-0.80), but I can make a diff against
older or debianized one.
--
WBR,
xrgtn
--- Linux-PAM-0.80/libpam_misc/misc_conv.c.orig 2005-10-30 21:54:33.000000000
+0200
+++ Linux-PAM-0.80/libpam_misc/misc_conv.c 2005-10-30 21:56:34.000000000
+0200
@@ -175,6 +175,7 @@
(void) tcsetattr(STDIN_FILENO, TCSAFLUSH, &term_tmp);
fprintf(stderr, "%s", prompt);
+ fflush(stderr);
if ( delay > 0 && set_alarm(delay, &old_sig) ) {
D(("<failed to set alarm>"));
