[settings] fixed: set timezone during initialization

[vkosh]

This sets timezone during xbmc loading.
To reproduce the bug:

1. go to settings and set timezone to something different from system one, note the time changed on home screen.
2. restart xbmc.
3. note that time on home screen corresponds to system timezone, but not the selected one.
4. go to settings and reset timezone, note the time changed to selected timezone.

[t-nelson]

Does this need to happen more generally to hit platforms besides linux?

[Montellese]

Using the OnSettingUpdate() is not the right way to go. The proper way to do this is to manually set the timezone on linux during startup after having loaded the settings. That's how it works for everything else as well (resolution, gui language etc).

@t-nelson: IIRC only linux has a timezone implementation right now.

[t-nelson]

@Montellese: OK, I couldn't remember if the other platforms diddled TZ or
not.

@vkosh: I was also questioning the implementation via the update callback.
After looking at the settings code again, I agree with @Montellese. TZ
should be set during CApp init.

[davilla]

darwin does.

[vkosh]

@t-nelson: I think CApplication::Initialize is also not correct place for setting TZ, because, for example, switching between profiles with the different TZ settings will not work in this case.
The right way IMO is to implement ISettingsHandler::OnSettingsLoaded in CLinuxTimezone.
@Montellese @t-nelson: if there is no objection I'll implement the fix this way.

[t-nelson]

As @davilla has pointed out, darwin also diddles TZ. So, CLinuxTimezone
isn't the right place, IMO.

[vkosh]

@t-nelson: As I can see in CSettings, CLinuxTimezone is the only handler for locale.timezone and locale.timezonecountry settings, and it affects linux systems only.

@davilla: Do the settings really change TZ on darwin platforms?

[vkosh]

Looks like it's another regression.
On Frodo we have initialization during settings loading and TZ settings should work on any "_LINUX" platform except android: see https://github.com/xbmc/xbmc/blob/Frodo/configure.in#L551 and https://github.com/xbmc/xbmc/blob/Frodo/xbmc/settings/GUISettings.cpp#L1424
On Gotham there is no initialization (subject of the PR) and "TARGET_LINUX" platforms supported only: see https://github.com/xbmc/xbmc/blob/master/m4/xbmc_arch.m4 and https://github.com/xbmc/xbmc/blob/master/xbmc/settings/Settings.cpp#L1142

So, I think the PR should be rewritten as I mentioned above. And the second PR should be created to support platforms that were supported in Frodo.

[vkosh]

Here is the new patch for tz initialization. I use TARGET_LINUX as the rest CLinuxTimezone callbacks currently registered for this target only. The other platforms support will be added in another PR.

[Montellese]

Yup looks like I missed the fact that darwin also handles timezones probably because of the name of CLinuxTimezone.

[vkosh]

@Montellese: thanks, fixed

[t-nelson]

jenkins build this please

[vkosh]

looks like a problem with jenkins

[t-nelson]

jenkins build this please

[t-nelson]

Manual droid86 build queued at http://jenkins.xbmc.org/job/XBMC-Android-X86/227/

[koying]

IIRC, TARGET_LINUX is true for android, too, and this seems to cause a regression there...
cf. http://forum.xbmc.org/showthread.php?tid=186294

[t-nelson]

I can't repro, but I have a test build going at
http://jenkins.xbmc.org/view/Dashboard/job/XBMC-Android-ARM/263/ with some
!defined(TARGET_ANDROID)s added.

I'll post the link in that thread when it's done.