add a hack to fix #64662. When the shell is something that we don't really

Mon Jul 22 16:45:43 2002  George Lebl <jirka@5z.com>

	* daemon/slave.c: add a hack to fix #64662.  When
	  the shell is something that we don't really know (we know
	  a few shells) just start the session with bash or sh (whichever
	  we find).

	* daemon/slave.c, daemon/server.c, daemon/gdm.[ch],
	  config/PreSession, config/PostSession: Write out an fake Xservers
	  file each time before the PreSession script so that we may use the
	  sessreg command nicely.  Also look for sessreg in the path and
	  a few extra dirs.  And use the utmp file as well.

	* config/Default: setup the system Xmodmap here, copied from kdm
	  a bit.  This should fix #86098, at least it does for me.

	* config/PreSession, daemon/slave.c: Run the PreSession script from
	  the session pid and in the PreSession script exec sessreg.  This
	  fixes the utmp bugs as sessreg will now register the correct pid
	  in utmp.  Also move the .xsession-errors before the PreSession
	  script and let through the stdout/stderr descriptors so that
	  we can get errors such as sessreg failing.  Also don't error out
	  when PreSession fails if this is a failsafe session.

	* daemon/slave.c: the pid of the session is not cleared until the
	  session is over.  It's not a race as I previously thought and in
	  fact it could leave processes behind maybe.

	* daemon/errorgui.c: make the errors display wider, recenter the
	  window and do a bit of cleanup

1 files changed, 3 insertions, 1 deletions

diff --git a/TODO b/TODO
index 2351a7cc..0345c621 100644
--- a/TODO
+++ b/TODO
@@ -1,4 +1,6 @@
-Postfreeze things:
+TODO things (some must wait until 2.1):
+  - check for /dev/null and make make sure we do all the open stuff correctly
+
   - handle kills of server not working correctly to be more robust
     we must leave killing to the main daemon since that can more likely kill
     the server.