[haiku-appserver] Re: [Haiku-commits] r12953 - haiku/trunk/src/servers/app
- From: Adi Oanca <adioanca@xxxxxxxxxxxxx>
- To: haiku-appserver@xxxxxxxxxxxxx
- Date: Mon, 06 Jun 2005 23:11:51 +0300
Axel Dörfler wrote:
> Adi Oanca <adioanca@xxxxxxxxxxxxx> wrote:
>
>>>+// NOTE: I added this here and it solves the invalid region
>>>problem
>>>+// for Windows that have been resized before they were shown. -
>>>Stephan
>>>+RebuildFullRegion();
>>>+SendViewCoordUpdateMsg();
>>>+
>>>+ if (invalidate)
>>> GetRootLayer()->GoInvalidate(this, fFull);
>>> }
>>
>> You should, under no circumstances call RebuildFullRegion() from
>>another thread than RootLayer's one. I can guarantee this can have
>>very
>>ugly consequences. For example try moving a window while you show/
>>hide
>>another.
>
> Okay, but how should it be done instead, then?
The 2 lines below Stephan's ones do the right thing. They send a
message to RootLayer's thread telling it what to invalidate.
Now that I look better, it looks that Stephan did not proceeded badly.
The first time when I read those lines I was under impression I saw
RebuildFullRegionS() which was very bad, but NOW I see the 's' is
absent, and that's a very good thing! :-)
No, the change is fine for the present clipping code. In fact I would
congratulate Stephan for pulling this off - it shows he understood the
clipping code. :-)
> Also, if the app_server crashes, it's almost always in the layer code -
> I don't see any obvious bugs in there, so it seems to be a locking
> problem.
True. There some small things left to be done. I know them, but if you
can tell me the situations in which these problems occur it would be of
help too.
> Those bugs were already there before Stephan's changes. Can
> you have a look at it when you're done with the clipping stuff?
Sure. It's next on my list.
bye,
Adi.
Other related posts:
