On Thu, 10 Mar 2011 12:00:33 +0100 Cedric BAIL <cedric.bail@...> said:
> On Thu, Mar 10, 2011 at 7:35 AM, Carsten Haitzler <raster@...>
> > On Wed, 2 Mar 2011 23:30:55 +0100 Andreas Volz <lists@...> said:
> >> Am Wed, 2 Mar 2011 00:22:56 +0100 schrieb Andreas Volz:
> >> > my application crashes after recompiling EFL:
> >> >
> >> > http://codepad.org/gQc8fqaL
> >> >
> >> > Reason is that elm_config=NULL while dereferencing and using the
> >> > macro ENGINE_COMPARE.
> >> >
> >> > Any ideas?
> >> I just found the "problematic" commit (rev. 56218): (raster)
> >> > dont ALWAYS init elm for externals - if one isnt used.. why init it?
> >> > (for that matter externals has a problem in design - it loads every
> >> > module for externals and thus links in every lib even if never used by
> >> > the app/edje obj).
> >> If I comment out e.g. the elm_shutdown() in external_elm_shutdown()
> >> (line 27) my application works again.
> >> I'll explain you my application setup. I start a plain Edje application
> >> without any linking or initialization to elementary. Then it loads user
> >> defined edj files. And in some of there are Elementary widgets. I
> >> couldn't know this before.
> >> Displaying my first edj file works great, but after deleting it I got
> >> this crash. This is problematic as the main application logic is to
> >> load and unload edj files...
> >> Do you've any ideas how to solve this problem? Could you maybe revert
> >> your patch until a final design solution is found. If not I've to
> >> maintain local patches for my project. :-(
> > actually i wrote a quick test that inits and shuts elm down - _elm_config is
> > not NULL. this is a bit trickier. i thought i'd be able to reproduce the
> > problem in this way... the issue is somewhere else and i'd like to be able
> > to reproduce it and fix it. right now "it works for me". test app attached.
> > (i did find valgrind complaining of some deferred window deleting handling
> > which i have now fixed).
> There was a bug in the refcount of elm external, that I fixed Tuesday.
> That's maybe why it work for you now :-)
oh well poopie then! at least i fixed another lurking issue :)
------------- Codito, ergo sum - "I code, therefore I am" --------------
The Rasterman (Carsten Haitzler) raster@...