[omniORB] Re: Testing OmniORB4 for memory leaks on Linux RH9

[Sean Parker]

Duncan - 

  Thanks - I'll try valgrind first before doing any more
code-dives. I'm new to Linux so any hints towards usage of
tools that others have used that give 'better' results are
appreciated. Of course, mtrace has its limitations, and I
suppose valgrind looks like its quite a bit more
substantial...

--- Duncan Grisby <duncan at grisby.org> wrote:
> 
> Some things allocated by static initialisers and by
> ORB_init cannot be
> safely deleted until the ORB library and all stubs are
> unloaded, so in
> some circumstances, you will still be able to see some
> "leaks" after
> orb->destroy(). 

Is this true even when using the non-shared archive
libraries? I'm not sure if mtrace has non-trivial issues
WRT statically allocated C++ objects

> If you repeatedly initialise and destroy,
> you shouldn't
> see any increase in the "leaks". valgrind on Linux
> doesn't report any
> memory leaks.
> 

10-4 - I'll try it out...

> But yes, if you find any leaks, I will welcome any
> patches that remove
> them.
> 
> Cheers,
> 
> Duncan.
> 
 thanks

   Sean





God Bless 
    Sean Parker 




__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com