Re: I think I found a freak bug on GTK+-1.3.15

[Adam <adam cfar umd edu>]

> This type of thing is very likely indicative of some memory corruption
> in your code.  It's unlikely to be in gtk (but could be).  Just the fact
> that removing the second pointer declaration makes it work indicates
> that it's your code that's messing up somewhere.  Just declaring a
> pointer doesn't do any object initialization (ie gtk_tree_view_new type
> stuff).  What it does do is change your stack layout slightly such that
> whatever memory corruption you're doing won't cause an immediate crash
> when you remove that unused variable.  Adding it back in changes offsets
> just slightly so that something important on the stack is overwritten.
>
> Pointer errors always maifest themselves in this type of manner.  Very
> frustrating.  Good luck.  You'll want to run your code through DDD
> debugger.  It has some very powerful facilities for viewing your
> variables etc.  You can then watch what happens when the offending line
> is executed.  Also you can load the core file into the debugger after
> the fact and see what line triggered the core dump, see the variables,
> etc.  Should help you establish what's happening.  (DDD is run by the
> command ddd.  It's downloadable from the internet too -- see
> freshmeat.net)

actually I would suggset memory debugger. some of those include

                efence
                njamd
                purify (non-free)

my favourie is njamd (it comes with redhat 7.2 and possibly others)

run it until it stop showing any errors and the above problem should
get auto-magically corrected

-- 
Adam
http://www.eax.com      The Supreme Headquarters of the 32 bit registers