[Qt-creator] Qt Creator 3.1 Slow Debugging on OS X 10.8

[Michael Jackson]

On May 14, 2014, at 6:22 PM, André Pönitz <apoenitz at t-online.de> wrote:

> On Wed, May 14, 2014 at 01:22:14PM -0400, Michael Jackson wrote:
>> I am trying out the QtCreator 3.1 release for OS X (10.8.5 with Xcode
>> 5.1.1 Tooling and Xcode 5.1.1 5B1008 Command line tools) and debugging is
>> extremely slow. By this I mean that I set a breakpoint in my code and it
>> hits that code quickly. But then getting the "Locals and Expressions" to
>> populate is another 15 seconds and then trying to scroll the list is
>> another 10 seconds to start. Disclosing the "this" variable on my QWidget
>> derived class takes another 10~20 seconds. THis is unbearable. I did not
>> have this issue with QtCreator 3.0.1 on the same system.  I believe I
>> have everything setup to use the LLDB on the "System" (/usr/bin/lldb). I
>> am debugging a self built Qt 4.8.5 application. This on a 2009 Mac Pro
>> with 8 Real Cores, 32GB of Ram and an SSD.
>> 
>> Is this normal for this system? Is it the LLDB support? Is anyone else
>> seeing this problem?
> 
> It's very likely the special support for QObject properties display. This
> needs calling function in the debugged program, which is inherently slow on
> any debugger backend. It's exceptionally bad for QWidget objects themselves
> (not so for derived classes), as there are a lot of "owned" properties.
> 
> Since I haven't actually benchmarked with Qt 4.8.5 this might even be worse
> than "expected", likely even completely failing and recovering only slowly.
> 
> 
> Could you please open the file share/qtcreator/debugger/lldbbridge.py of
> your Qt Creator installation in a text editor and change the line that reads
> 
>    staticMetaObject = self.extractStaticMetaObject(value.GetType())
> 
> by
> 
>    staticMetaObject = None
> 
> and check whether this makes a difference?
> 
> That should be line 1028 in that file, or close to it.
> 
> 
> Andre'

Thanks for the tip. This does return the debugging speed back to acceptable. And it seems like I am getting most of the variables and variable values.

It was line 1037.

Thanks
Mike Jackson