Solid upower backend and kdeinit crash...

[Dawit A]

On Fri, Oct 29, 2010 at 3:13 PM, Chani <chanika at gmail.com> wrote:
> On October 29, 2010 19:48:49 Dawit A wrote:
>> Hi,
>>
>> The upower backend, more specifically the XRandrBrightness class,
>> causes kdeinit to crash on my system when you come back from screen
>> lock mode. This happens because when the check for the backlight
>> property fails in the constructor of XRandrBrightness, which happens
>> on my system, it returns immediately. That leaves the m_resources
>> variable set to NULL. Unfortunately both the brightness and
>> setBrightness member functions use m_resources without even bothering
>> to check whether it is null or not and hence the crash. Actually, the
>> call to XRandrBrightness::brightness can be traced one step back to
>> PowerDevilUPowerBackend::brightness which also does not bother to
>> check whether the action it wants to obtain is supported or not before
>> calling it, i.e. it does not do  XRandrBrightness::isSupported...
>>
>> Anyhow, the issue can be fixed in a couple of ways, not returning when
>> backlight check fails in ctor or checking validity of the m_resources
>> pointer before using it. Not sure which would be the correct fix
>> though. Any ideas ?
>
> checking pointers (and supportedness) is generally a good idea. :)
>
> but... it returns in the *constructor*? that sounds a little odd.

Yes, it does....

> I hope the contructor has at least initialized all its variables by then, so that you
> know m_resources will be null and not random garbage.

Yes, it does initialize m_reources to null.

> I'd probably have to read the code to know whether it was sane (and I can't tell what constructor
> you're referring to)

kdebase/workspace/powerdevil/daemon/backends/upower/xrandrbrightness.cpp