The non empty patch.
Peter Kümmel wrote:
> Enrico Forestieri wrote:
>> On Wed, Dec 20, 2006 at 07:57:30PM +0100, Peter Kümmel wrote:
>>> Bo Peng wrote:
>>>> It appears to me that ~LyxView() is called before ~Buffer() so when
>>>> the closing() signal is emitted (line 229, buffer.C), and triggers a
>>>> LyXView::setBuffer, there is no instance of LyXView. This is the also
>>>> the case for msvc, but lyx/win does not report any error.
>>>>
>>>> Peter, does this have anything to do with your lyx/quit cleanup?
>>>>
>>>> Bo
>>>>
>>> Enrico was the last one who touched the quit logic.
>> Yes, but I only made sure that the quit procedure was not entered at
>> all if cancel was selected.
>>
>
>
> lyx_main.C:
>
> /// The main application class private implementation.
> struct LyX::Singletons
> {
> Singletons(): iconv(ucs4_codeset, "UTF-8")
> {
> }
> /// our function handler
> LyXFunc lyxfunc_;
> ///
> BufferList buffer_list_;
> ///
> boost::scoped_ptr<kb_keymap> toplevel_keymap_;
>
>
> buffer_list_ is not in a smart pointer, don't know why.
> And there is no explicit destruction of the Singleton
> object, so the compiler will do it.
>
> Maybe attached patch helps.
>
>
--
Peter Kümmel
Index: src/lyx_main.C
===================================================================
--- src/lyx_main.C (revision 16341)
+++ src/lyx_main.C (working copy)
@@ -181,6 +181,7 @@
LyX::~LyX()
{
+ pimpl_.reset();
}
