Do you run valgrind on juce apps?

General JUCE discussion
1

Hi all,

I tried to run valgrind to check for memory problems on a small test project, before integrating it to the main project
It’s the first time I run valgrind on a Juce project, and valgrind reports plenty of problems. Well that’s not a huge surprise. What is surprising is where the problems appear.

Have a look at the first reported error for exemple :


==9892== Invalid write of size 8
==9892==    at 0x47F608: juce::ScopedPointer<juce::DocumentWindow::ButtonListenerProxy>::ScopedPointer() (juce_ScopedPointer.h:58)
==9892==    by 0x47D6FE: juce::DocumentWindow::DocumentWindow(juce::String const&, juce::Colour const&, int, bool) (juce_DocumentWindow.cpp:79)
==9892==    by 0x4108E8: MainAppWindow::MainAppWindow() (MainWindow.cpp:19)
==9892==    by 0x40A99D: TestApplication::initialise(juce::String const&) (Main.cpp:40)
==9892==    by 0x415EAF: juce::JUCEApplication::initialiseApp(juce::String const&) (juce_Application.cpp:171)
==9892==    by 0x416174: juce::JUCEApplication::main(juce::String const&) (juce_Application.cpp:223)
==9892==    by 0x41637C: juce::JUCEApplication::main(int, char const**) (juce_Application.cpp:263)
==9892==    by 0x408D3F: main (Main.cpp:132)

It’s kind of confusing because my own code here, is only creating a mainAppWindow here !
On the other hand, it’s hard for me to believe that there are memory errors in a library like juce. Furthermore, I’ve seen times, for example, whith visual studio leak checker, where a leak is reported in Juce, but caused by the calling code .

Well in the end, what I would like to know, is whther jules, or someone else, tried to use valgrind on code using juce, and it came clean ?

I don’t really know how to interpret the error above (I’ve quite a few like this)

  • Is it a false positive ?
  • Is it a real error, but the reported location is not where the cause of the error lies ?
  • There really are memory problems inside juce ? :slight_smile:

What do you guys think ?

P.S. : The exact same code, ran on windows leads to no error at all, with all the possible run-time checks enabled in visual studio.

2

I do occasionally run valgrind as a quick check, but never saw that happen - I’ll have a go myself and see if I can spot any problems. That error looks very odd though, it seems to be complaining about the compiler-generated constructor code for ScopedPointer, which really can’t be wrong… maybe some kind of link or compile problem?

3

Yes, it’s very strange, I’ll have a look on my side too. Thanks for checking.