[ros-dev] ros-amd64-bringup! Surprise! No Region Leaks!
jimtabor.rosdev at gmail.com
Fri Feb 5 13:22:36 CET 2010
You maybe right.
We are still looking into the logs here. The last thing I did and
looks like the only commit was late November working on getting
Classes fixed. Bug 4980 was filed on the 2nd of December. So this
means the leak has been around for a while. FF 3.5 and SM 2.0 comes
out about this time too. Remembering back from the Testers paste posts
about GDI object crashes during web surfing. At the time, I was busy
working on Classes and user32. There was a bug filed about AbiWord
with a funny combo box issue when selecting the font type, then
isolated it to an ARM3 kernel commit. Remembering again, that was
about in June/July when I noticed it after upping and thinking I had
done it. After reverting, I knew I was not the one and moved on. I did
look to see if one of the wine sync mergers created the leak and I
followed the flow and the regions had been deleted. Just glancing
through the files......
So~ That leak must have been around for a while. We had two releases
so my guess is to have the testers have a crack at it and see. But,!!,
FF 3.5 and SM 2.0 will not run in those releases....... Best test I
have to check, run GDIviewer with wordpad. Running wordpad, select one
of the menus and run the mouse over to the next menu and observe the
Region count in GDIviewer. GDIviewer can auto refresh so play with it.
Why FF 3.5 and SM 2.0? These are the worse case tests, giving the time
to fill up the processes object counts, very fast!
On Fri, Feb 5, 2010 at 5:46 AM, Timo Kreuzer <timo.kreuzer at web.de> wrote:
> James Tabor schrieb:
>> Timo told me. So it must build in 32 bit mode and boot.
> Oops my fault, I guess :)
> I thought you were joking when talking about the branch. That's why I
> said it doesn't have the kernel problems of trunk and there is no region
> leak. Because the 64 bit code it is neither subject to trap handling
> rewrite and of cause does not boot anywhere near that. I didn't take it
> as you wanted to use the 32bit compiled version.
> It should boot when compiled for 32 bit, but I don't actually know
> whether the leak is there or not, I didn't check yet.
More information about the Ros-dev