Commit 7b159b55d6398d2d240817276ae27ed68f11deda breaks ABI?

[Matthew Garrett]

On Fri, Aug 31, 2007 at 10:01:11PM +0200, Danny Kukawka wrote:

> The problem is/was: the scripts return UINT32 to the caller, but the addons as 
> e.g. the macbook or dell brightness addon return in. And if someone check the 
> return value for GetBrightness() it differ between the scripts and the 
> addons, but the caller as to assume he get always the same return type from 
> the same interface. If you only reverted the commit IMO/IIRC this would break 
> again all returns from the addons. 

The spec explicitly stated that they returned an int at that point, not 
a uint, and as far as I can tell from the commitdiff that's what they 
did - the commit changed them to UINT32.

-                dbus_message_iter_append_basic (&iter, DBUS_TYPE_INT32, &result);
+                dbus_message_iter_append_basic (&iter, DBUS_TYPE_UINT32, &result);



> And if I understand the comment in [1] and [2] correct it would break also 
> g-p-m (as now again KPowersave and powersaved) or not?

g-p-m was broken to begin with. The problem now is that it's impossible 
to produce a version of g-p-m that works with both 0.5.9 (the latest 
release of hal) and the release candidate. The only way to avoid this is 
to ensure 0.6.0 has the same behaviour as 0.5.9 did.

-- 
Matthew Garrett | mjg59 at srcf.ucam.org