Changes have been made in Git (FRAMEWORK_5_0):

commit 1fa6e80cc8ea4849d5cdf4857c22c8e88417476d
Author: Thomas Jarosch <thomas.jarosch@intra2net.com>
Date:   Fri Apr 19 17:16:23 2013 -0400

     Correctly set utcduedate and duedate properties.

     Bug: 12200
     Signed-off-by: Michael J Rubinsky <mrubinsk@horde.org>

  nag/lib/Task.php |    4 ++--
  1 files changed, 2 insertions(+), 2 deletions(-)

http://git.horde.org/horde-git/-/commit/1fa6e80cc8ea4849d5cdf4857c22c8e88417476d

Committed, thanks!

Explanation makes perfect sense, I was getting confused by the fact 
that the documentation says all dates must be sent in UTC.

Changes have been made in Git (master):

commit da4b863c5fc517c0e6c8cca283a4c89910f0edb2
Author: Thomas Jarosch <thomas.jarosch@intra2net.com>
Date:   Fri Apr 19 17:16:23 2013 -0400

     Correctly set utcduedate and duedate properties.

     Bug: 12200
     Signed-off-by: Michael J Rubinsky <mrubinsk@horde.org>

  nag/lib/Task.php |    4 ++--
  1 files changed, 2 insertions(+), 2 deletions(-)

http://git.horde.org/horde-git/-/commit/da4b863c5fc517c0e6c8cca283a4c89910f0edb2

So, just keep doing it like I have been then.
The patch gets applied to the tree as the person who submitted it. The 
person from Horde who commits it (and therefore approves it) is the 
one who "signed off" on it. This gives the author credit (his name is 
in the repository) and the person who approved it can be then be later 
tracked down and beaten with a wet noodle if the patch broke something 
important.

Either remove the Signed-off-by stuff completely if it's unwanted by horde
or just add yours without replacing mine :)

The linux kernel subsystem maintainers add their Signed-off-by next to 
the original author's one, but I guess horde does not need the strong 
"guarantees" as the kernel people do.

This one explains it quite good:
http://stackoverflow.com/questions/1962094/what-is-the-sign-off-feature-in-git-for

-> Probably it's easier if I just generate patches without that stuff ;)

Here's an example from the WP8 phone using the fixed code:

---------------------
2013-04-19T17:56:17+02:00 DEBUG: [13545] I       <POOMTASKS:DueDate>
2013-04-19T17:56:17+02:00 DEBUG: [13545] I         2013-04-28T00:00:00.000Z
2013-04-19T17:56:17+02:00 DEBUG: [13545] I        </POOMTASKS:DueDate>
2013-04-19T17:56:17+02:00 DEBUG: [13545] I       <POOMTASKS:UtcDueDate>
2013-04-19T17:56:17+02:00 DEBUG: [13545] I         2013-04-27T22:00:00.000Z
2013-04-19T17:56:17+02:00 DEBUG: [13545] I        </POOMTASKS:UtcDueDate>
---------------------

The "DueDate" and "UtcDueDate" sent by horde without the fix were 
exactly the same values. On all tested phones this caused the due date 
to be off by two hours for my timezone, so the due date jumped to the 
previous day.

I _think_ (I haven't read the spec) that "DueDate" is in the current 
timezone, otherwise it would not make sense to have the additional 
"UtcDueDate" field.

Sure, not a problem, though I've been committing them as authored by 
you, with me signing off on it. You want it authored by you and you 
signing off on it too?

Can you explain the symptoms on the device that this was causing? When 
I wrote this I was confused by the two fields because of how all dates 
transmitted to the client must be in UTC anyway, and none of the 
clients I have that support tasks seem to behave any differently with 
those different values.

Patch to fix the issue.

Please keep my "Signed-off-by:" line when applying the patch,
just append yours below mine.

Hi,

the "due by" and "delayed start" fields currently sent to the AS device
are set to the same value. Probably only works if you are living in GMT ;)

Attached patch fixes the issue.

Tested with:
- HTC Desire HD / Android 4.04
- Nokia N900 / maemo 5
- Nokia Lumia 920 / Windows phone 8

Thomas