This looks like a good start.

@bu5hm4n, @ProhtMeyhet are you happy with the rebased version now?

Take comments into account.

Thanks for taking the time and making this usable! :-)

I confirm this patch removes the C++17 warnings I reported, but I'll let somebody else review the code.

@segfaultxavi D11638 should deal with it.

That does also not work, because a eventual execp would overwrite anything in there, that is not in a environmemt variable .

In that case you should probably have a ecore_fork_reset_callback_add() call to correctly handle forks.

@zmike mhm it cannot fix the issue here, as the same might happen theoretically due to evas forking, or efl_net or eio or eeze ... we have a lot places where we do that ...

efreet_cache_disable() for exactness overall might be a more permanent solution here.

I confirm this fixes my problems with Exactness. Please land.

this time a working possibility

Thanks guys.

@Jaehyun Is that a feature or a miracle? It works here. @bu5hm4n probably a good idea, the solution is a bit tricky and could do with being documented.

I think that if EDI's wizard registers pop_cb and does nothing in pop_cb and returns EINA_FALSE in the pop_cb, then the last item is not popped.

While i agree with @Jaehyun_Cho what do you think of adding a flag to enable that feature ? The feature itself sounds usefull to me :)

I think this patch breaks the backward compatibility.
Because users can register their own pop_cb by using elm_naviframe_item_pop_cb_set().
So if the popped item is the last item, then users' pop_cb is called and elm_win or something is deleted manually in the pop_cb.
However, this patch do not allow pop_cb is called if the popped item is the last item.
So I think this patch should not be accepted.

git add -p mistake

@devilhorns thanks for your review, but I would like to abandon this patch.