seg fault on master

[John Ralls]

> On Jul 5, 2015, at 10:27 PM, John Ralls <jralls at ceridwen.us> wrote:
> 
> 
>> On Jul 5, 2015, at 6:43 PM, Alex Aycinena <alex.aycinena at gmail.com> wrote:
>> 
>> Program received signal SIGSEGV, Segmentation fault.
>> 0x00007ffff2025cf2 in std::_Rb_tree<char const*, std::pair<char const* const, KvpValueImpl*>, std::_Select1st<std::pair<char const* const, KvpValueImpl*> >, KvpFrameImpl::cstring_comparer, std::allocator<std::pair<char const* const, KvpValueImpl*> > >::_M_lower_bound (this=0x32069c0, __x=0x31, __y=0x3a286f8, 
>>    __k=@0x7fffffffbda0: 0x33c16b8 "Budgeting")
>>    at /usr/include/c++/4.9.2/bits/stl_tree.h:1277
> 
> This one is harder. I’ve found the cause, but it’s bedtime so I’ll have figure out what to fix tomorrow. The fundamental problem is that when I call “new KvpFrame” in sixtp-dom-parser.cpp, it’s allocating memory that libxml2 thinks belongs to it, so when the “data_for_children()” subtree is freed in gnc_book_slots_end_handler(), the frame belonging to that KvpValue gets freed and later overwritten by a call to g_log. When the optiondb tries to access later, boom.

Alex,

Well, that was the wrong cause. The real one was that I didn’t delete the option slots the right way. That helped me notice that KvpFrameImpl needed a proper destructor. Both committed, book options work again.

Regards,
John Ralls