opening parallel tab a second time crashes

[karlkleinpaste]

Reproducible every time: Start Xiphos, open parallel tab, close it, open it again, crash.

(xiphos:2242593): GLib-GObject-CRITICAL **: 08:55:15.399: invalid unclassed pointer in cast to 'GtkWidget'
(xiphos:2242593): Gtk-CRITICAL **: 08:55:15.399: gtk_widget_get_realized: assertion 'GTK_IS_WIDGET (widget)' failed
Thread 1 "xiphos" received signal SIGSEGV, Segmentation fault.
0x00007ffff1d35228 in g_type_check_instance_cast () from /lib64/libgobject-2.0.so.0
(gdb) bt 10
#0  0x00007ffff1d35228 in g_type_check_instance_cast () at /lib64/libgobject-2.0.so.0
#1  0x000000000048fe11 in main_update_parallel_page_detached () at /home/karl/src/bible/xiphos/src/main/parallel_view.cc:952
#2  0x000000000043a774 in on_entry_activate (entry=0xb096810, user_data=0x0) at /home/karl/src/bible/xiphos/src/gtk/navbar_versekey_parallel.c:328
#3  0x00007ffff1d0da2c in g_closure_invoke () at /lib64/libgobject-2.0.so.0
#4  0x00007ffff1d2baba in signal_emit_unlocked_R.isra.0 () at /lib64/libgobject-2.0.so.0
#5  0x00007ffff1d2daf6 in signal_emit_valist_unlocked () at /lib64/libgobject-2.0.so.0
#6  0x00007ffff1d2dd68 in g_signal_emit_valist () at /lib64/libgobject-2.0.so.0
#7  0x00007ffff1d2de23 in g_signal_emit () at /lib64/libgobject-2.0.so.0
#8  0x00007ffff18ebe4c in gtk_widget_activate () at /lib64/libgtk-3.so.0
#9  0x00000000004a9cd9 in show_parallel (svalue=0xe872b70 "1Cor 8:7", stype=0x59f3660 "verse", clicked=1) at /home/karl/src/bible/xiphos/src/main/url.cc:271
(More stack frames follow...)
(gdb) up
#1  0x000000000048fe11 in main_update_parallel_page_detached () at /home/karl/src/bible/xiphos/src/main/parallel_view.cc:952
952		    !gtk_widget_get_realized(GTK_WIDGET(widgets.html_parallel_dialog)))
(gdb) l
947		SWBuf text("");
948		gchar buf[5000];
949		gint modidx, parallel_count, fraction;
950	
951		if (!widgets.html_parallel_dialog ||
952		    !gtk_widget_get_realized(GTK_WIDGET(widgets.html_parallel_dialog)))
953			return;
954	
955		/* how big a pile of parallels have we got? */
956		if (settings.parallel_list == NULL)
(gdb) p widgets
$1 = {app = 0x5f5830, appbar = 0x5ec430, vboxMain = 0xc6fbe0, page = 0xc39010, epaned = 0x235d9f0, paned_sidebar = 0x58f2520, box_side_preview = 0x58f2c60, vpaned = 0x25e71e0, vpaned2 = 0x3253670, 
  hpaned = 0x2da3b50, bible_info_paned = 0x0, shortcutbar = 0x58f2520, bookmark_tree = 0x5917720, ctree_widget_books = 0x0, workbook_lower = 0x0, notebook_gbs = 0x0, notebook_dict = 0x0, 
  all_strongs = 0x23602f0, box_comm = 0x303d5c0, notebook_bible_parallel = 0x27837b0, notebook_text = 0x28b18b0, notebook_percomm = 0x0, notebook_sidebar = 0x58fb890, notebook_comm_book = 0x243dd70, 
  notebook_main = 0x243f430, notebook_dict_devot = 0x2951530, vbox_text = 0x3132640, box_dict = 0x276a880, vbox_previewer = 0x2c1cb90, parallel_tab = 0xe88b110, dock_sb = 0x0, frame_parallel = 0x0, 
  studypad_dialog = 0x0, entry_dict = 0x27637d0, toolbar_comments = 0x0, toolbar_studypad = 0x0, eventbox1 = 0x0, html_search_report = 0x0, html_parallel = 0xac6a860, 
  html_parallel_dialog = 0x4ba3690, html_text = 0x31317d0, html_comm = 0x2dde300, html_percomm = 0x0, html_dict = 0x2aad340, html_devotional = 0x4a68c70, html_book = 0x273dd30, html_studypad = 0x0, 
  html_previewer = 0x0, html_previewer_text = 0x29a0bc0, cb_books = 0x0, listview_dict = 0x0, store = 0x0, label_comm = 0x0, label_dict = 0x0, comboboxentry_dict = 0x0, entry_devotional = 0x2364fa0, 
  button_devotional_date = 0x30e7ce0, popover_devotional = 0x2f30cc0, calendar_devotional = 0x2f4e230, add_bookmark_submenu = 0x0, viewtexts_item = 0x2e6a470, viewpreview_item = 0x3169ed0, 
  parallel_tab_item = 0x3225690, side_preview_item = 0x3226460, viewcomms_item = 0x318ee40, viewdicts_item = 0x339f5c0, linkedtabs_item = 0x2622140, showversenum_item = 0x2623a90, 
  readaloud_item = 0x2622da0, versehighlight_item = 0x29aba20, annotate_highlight_item = 0x29aca10, xrefs_in_verse_list_item = 0x0, new_journal_item = 0x33287b0, button_dict_book = 0x0, 
  button_new_tab = 0x3203a50, button_dict_back = 0x2d2c3b0, button_dict_forward = 0x2409730, hboxtb = 0x3030b40}
(gdb) p widgets.html_parallel
$2 = 0xac6a860
(gdb) p widgets.html_parallel_dialog
$3 = 0x4ba3690
(gdb) p *(widgets.html_parallel)
$4 = {parent_instance = {g_type_instance = {g_class = Python Exception <class 'gdb.error'>: No type named TypeNode.
}, ref_count = 1, qdata = 0xac74a70}, priv = 0xac6a770}
(gdb) p *(widgets.html_parallel_dialog)
$5 = {parent_instance = {g_type_instance = {g_class = <error reading variable: Cannot access memory at address 0x140>}, ref_count = 32, qdata = 0x4ba7d13}, priv = 0x333a30322073}

Why does it test html_parallel_dialog when html_parallel should be what's under consideration for parallel tab? It's not opening the separate "dialog" window, it's opening the tab. No wonder that what's in the "dialog" structure isn't reliable.

Well...

git blame says I wrote those few lines of code (951-953) in 2012. But I never personally use the parallel tab, I use only the detached parallel dialog, so I've never noticed the problem. Question is, Why/how does opening the parallel tab lead to trying to update the parallel dialog? Maybe more importantly, why is this happening just now, with no user reports? Have recent updates induced the problem?

Something has (recently?) chosen to take the route through the dialog code when it should not even think to do so.

The parallel dialog can be opened and re-opened repetitiously without incident.

[LAfricain]

I think I managed to fix the crash, but now I have two navigation bars again! It's driving me crazy!! But since the crash is a bigger issue, I'm going ahead and pushing the changes now. I'll take another look at the double navigation bar later.

[LAfricain]

Sorry, I didn't apply the changes correctly.

[karlkleinpaste]

Appears fixed by #1373.

[karlkleinpaste]

I have to re-open this. The fix above is applied, but the crash is still happening...sometimes.

0x00007ffff1d35228 in g_type_check_instance_cast () from /lib64/libgobject-2.0.so.0
(gdb) bt 12
#0  0x00007ffff1d35228 in g_type_check_instance_cast () at /lib64/libgobject-2.0.so.0
#1  0x000000000048fe21 in main_update_parallel_page_detached () at /home/karl/src/bible/xiphos/src/main/parallel_view.cc:952
#2  0x000000000043a785 in on_entry_activate (entry=0x4b70380, user_data=0x0) at /home/karl/src/bible/xiphos/src/gtk/navbar_versekey_parallel.c:331
#3  0x00007ffff1d0da2c in g_closure_invoke () at /lib64/libgobject-2.0.so.0
#4  0x00007ffff1d2baba in signal_emit_unlocked_R.isra.0 () at /lib64/libgobject-2.0.so.0
#5  0x00007ffff1d2daf6 in signal_emit_valist_unlocked () at /lib64/libgobject-2.0.so.0
#6  0x00007ffff1d2dd68 in g_signal_emit_valist () at /lib64/libgobject-2.0.so.0
#7  0x00007ffff1d2de23 in g_signal_emit () at /lib64/libgobject-2.0.so.0
#8  0x00007ffff18ebe4c in gtk_widget_activate () at /lib64/libgtk-3.so.0
#9  0x00000000004a9ce9 in show_parallel (svalue=0xe61c3f0 "Song 5:4", stype=0x2369370 "verse", clicked=1) at /home/karl/src/bible/xiphos/src/main/url.cc:271
#10 0x00000000004ac5e9 in main_url_handler (url=0xe7276c0 "passagestudy.jsp?action=showParallel&type=verse&value=Song+5%3A4", clicked=1) at /home/karl/src/bible/xiphos/src/main/url.cc:1105
#11 0x000000000043c909 in sync_with_main (key=0x0) at /home/karl/src/bible/xiphos/src/gtk/parallel_tab.c:133
(More stack frames follow...)
(gdb) up
#1  0x000000000048fe21 in main_update_parallel_page_detached () at /home/karl/src/bible/xiphos/src/main/parallel_view.cc:952
952		    !gtk_widget_get_realized(GTK_WIDGET(widgets.html_parallel_dialog)))
(gdb) l
947		SWBuf text("");
948		gchar buf[5000];
949		gint modidx, parallel_count, fraction;
950	
951		if (!widgets.html_parallel_dialog ||
952		    !gtk_widget_get_realized(GTK_WIDGET(widgets.html_parallel_dialog)))
953			return;
954	
955		/* how big a pile of parallels have we got? */
956		if (settings.parallel_list == NULL)
(gdb) p *widgets.html_parallel_dialog
$1 = {parent_instance = {g_type_instance = {g_class = <error reading variable: Cannot access memory at address 0x189>}, ref_count = 65, qdata = 0x4b093f0}, priv = 0xe724400}

[karlkleinpaste]

Should we simply be ensuring that widgets.html_parallel_dialog is NULLed out? In my last testing, I hadn't invoked detached parallel dialog at all, so that part of the structure should never have been touched in the first place; thus, it must hold unreferenceable garbage.

[karlkleinpaste]

Looking really closely now for the first time, my first horrified realization is html_parallel_dialog is used for both the detached parallel dialog and for the embedded tab. html_parallel is for the 1-verse parallel alternate to the main Bible tab, refreshed with main_update_parallel_page().

main_update_parallel_page_detached() is used for the full chapter parallel, whether detached dialog or embedded tab.

Assuming I'm reading this right in the first place...

In my defense, I didn't write any of the parallel code.

[karlkleinpaste]

fixed by #1376.