summaryrefslogtreecommitdiff
path: root/TODO
diff options
context:
space:
mode:
authorBrian Cameron <Brian.Cameron@Sun.Com>2005-03-07 20:40:10 +0000
committerBrian Cameron <bcameron@src.gnome.org>2005-03-07 20:40:10 +0000
commitac17073801698dd6a30d4d8fc069d5df73ea4204 (patch)
treed7ba1eb59665bbc0ccbe5c6e5865f53e88b10800 /TODO
parent45d7b0c86051caad193f8355acb49c21e6be5567 (diff)
downloadgdm-ac17073801698dd6a30d4d8fc069d5df73ea4204.tar.gz
Release 2.6.0.8 Updated version. Corrected spelling and use betterGDM2_2_6_0_8
Mon Mar 07 14:35:00 2005 Brian Cameron <Brian.Cameron@Sun.Com> * Release 2.6.0.8 * configure.in: Updated version. * NEWS, ChangeLog, README, README.install, HACKING, TODO, daemon/auth.c daemon/errorgui.c, daemon/gdm.c daemon/slave.c, daemon/verify-pam.c, daemon/xdmcp.c, gui/gdmcomm.c, gui/gdmlogin.c, gui/gdmsetup.c, gui/greeter/greeter_action_language.c: Corrected spelling and use better english.
Diffstat (limited to 'TODO')
-rw-r--r--TODO67
1 files changed, 38 insertions, 29 deletions
diff --git a/TODO b/TODO
index c4ebea76..6d4b97be 100644
--- a/TODO
+++ b/TODO
@@ -1,40 +1,49 @@
-Big TODO: Rewrite the slave to be based around a mainloop. So that
-we don't have to do things directly from signal handlers. Some work has already
-been started by making the slave use select calls instead of waitpid. This needs
-to be extended a bit. The big issue is things that hang. We need two things,
-some better general framework for things that hang and handling signals there
-(see daemon/misc.c for solution when using gethostbyaddr in the main daemon).
-We also need some way to do something like "sleep(5)" with the mainloop nicely.
-We want the slave to still be very synchroneous, but just have some small handlers
-run from within the mainloop. The one thing that will always cause trouble here
-is the XDMCP server pinging, I think in this case it would be sane to do the setjmp
-thing and just ignore the ping if a signal comes, not sure how this will play with
-the X stuff and if this would completely whack us out. The other option is to have
-a separate "pinger process" but that seems to heavyweight. Another option here
-is to give up on complete mainloopishness and during pinging set up an alarm or
-some other polling thing that checks all the mainloops stuff.
-
-All the GUIs running as the gdm user should have some resource limits set to
-make it hard to do DoS attacks by somehow exploiting a leak or some such.
+
+gdm TODO list:
+
+Big TODO: Rewrite the slave to be based around a mainloop.
+So that we don't have to do things directly from signal handlers.
+Some work has already been started by making the slave use select
+calls instead of waitpid. This needs to be extended a bit. The
+big issue is things that hang. We need two things, some better
+general framework for things that hang and handling signals there
+(see daemon/misc.c for solution when using gethostbyaddr in the
+main daemon). We also need some way to do something like
+"sleep(5)" with the mainloop nicely. We want the slave to still
+be very synchronous, but just have some small handlers run from
+within the mainloop. The one thing that will always cause
+trouble here is the XDMCP server pinging, I think in this case
+it would be sane to do the setjmp thing and just ignore the ping
+if a signal comes, not sure how this will play with the X stuff
+and if this would completely whack us out. The other option is
+to have a separate "pinger process" but that seems to
+heavyweight. Another option here is to give up on complete
+mainloopishness and during pinging set up an alarm or some other
+polling thing that checks all the mainloops stuff.
+
+All the GUIs running as the gdm user should have some resource
+limits set to make it hard to do DoS attacks by somehow exploiting
+a leak or some such.
Small TODO things:
- - If we can't setup pam display user visible errors and not just syslog stuff
+ - If we can't setup pam display user visible errors and not just
+ syslog stuff
-Languages should be read from available locales. Look at source for
-locale_config. Currently the way we do things is utter shit.
+Languages should be read from available locales. Look at source
+for locale_config. Currently the way we do things is utter crack.
Perhaps stuff:
- Keyboard layout menu.
- native utmp/wtmp handling without sessreg
- xsri like logo functionality
-the photosetup proggie should be put into some sort of crapplet and there
-should be other settings in the crapplet as well.
+The photosetup proggie should be put into some sort of capplet and
+there should be other settings in the capplet as well.
+
+Handle all the XFree86 Xserver error thingies, such as No "Display"
+subsection for etc...
-handle all the XFree86 Xserver error thingies, such as
-No "Display" subsection for
-etc...
+Add spacer files. Files which we can use in case there is disk
+shortage to write a new xauthority stuff so that someone can actually
+log on and clean up. Is this crack? Maybe.
-Add spacer files. Files which we can use in case there is disk shortage to
-write a new xauthority stuff so that someone can actually log on and clean up.
-Is this crack? Maybe.