Re: EntryCompletion woes (maybe?)

[muppet <scott asofyet org>]

On Nov 7, 2006, at 7:09 PM, Matthew Braid wrote:

> Unfortunately there's been some strange behaviour since I added it  
> - namely if I try to delete a row _after text has been entered into  
> a combo box entry_ I get:
>
> Gtk-CRITICAL **: gtk_entry_get_text: assertion `GTK_IS_ENTRY  
> (entry)' failed at -e line 1.
>
> which I've tracked down to the EntryCompletions.

By setting --g-fatal-warnings and running perl in a debugger, i get  
this backtrace at the first of the CRITICAL messages:

#0  0x008ab7a2 in _dl_sysinfo_int80 () from /lib/ld-linux.so.2
#1  0x008eae59 in raise () from /lib/tls/libc.so.6
#2  0x008ec882 in abort () from /lib/tls/libc.so.6
#3  0x00e5b38e in gperl_log_handler (log_domain=0xbebebb "GLib- 
GObject", log_level=0, message=0x9dbfdc "<½\235", user_data=0x0) at  
GLog.xs:140
#4  0x0016cc44 in IA__g_logv (log_domain=0xbebebb "GLib-GObject",  
log_level=G_LOG_LEVEL_WARNING, format=0xbef780 "%s: instance `%p' has  
no handler with id `%lu'",
    args1=0xbfee2f8c "Aì¾") at gmessages.c:474
#5  0x0016cdcc in IA__g_log (log_domain=0xbebebb "GLib-GObject",  
log_level=G_LOG_LEVEL_WARNING, format=0xbef780 "%s: instance `%p' has  
no handler with id `%lu'")
    at gmessages.c:517
#6  0x00bdb8cc in IA__g_signal_handler_block (instance=0x8aef510,  
handler_id=281) at gsignal.c:1693
#7  0x00f46f4d in IA__gtk_entry_completion_insert_prefix  
(completion=0x8b490f0) at gtkentrycompletion.c:1599
#8  0x00f440ee in check_completion_callback (completion=0x8b490f0) at  
gtkentry.c:5382
#9  0x00bdf68d in source_closure_marshal_BOOLEAN__VOID (closure=0x6,  
return_value=0xbfee30b0, n_param_values=0, param_values=0x0,  
invocation_hint=0x0, marshal_data=0x0)
    at gsourceclosure.c:81
#10 0x00bce14e in IA__g_closure_invoke (closure=0x8b8e520,  
return_value=0xbfee30b0, n_param_values=0, param_values=0x0,  
invocation_hint=0x0) at gclosure.c:490
#11 0x00bdf7c0 in source_closure_callback (data=0x8b8e520) at  
gsourceclosure.c:123
#12 0x001687a4 in g_idle_dispatch (source=0x8b8f238, callback=0,  
user_data=0x8b8e520) at gmain.c:3796
#13 0x00165b8c in g_main_dispatch (context=0x891b188) at gmain.c:1916
#14 0x00166b51 in IA__g_main_context_dispatch (context=0x891b188) at  
gmain.c:2466
#15 0x00166e76 in g_main_context_iterate (context=0x891b188, block=1,  
dispatch=1, self=0x88c6548) at gmain.c:2547
#16 0x00167372 in IA__g_main_loop_run (loop=0x8b8cdb0) at gmain.c:2751
#17 0x00fad8eb in IA__gtk_main () at gtkmain.c:1000
#18 0x002b3c32 in XS_Gtk2_main (my_perl=0x8855008, cv=0x8929d1c) at  
xs/Gtk2.c:387
#19 0x0076326d in Perl_pp_entersub () from /usr/lib/perl5/5.8.3/i386- 
linux-thread-multi/CORE/libperl.so
#20 0x00745bbd in Perl_runops_debug () from /usr/lib/perl5/5.8.3/i386- 
linux-thread-multi/CORE/libperl.so
#21 0x006f4958 in perl_run () from /usr/lib/perl5/5.8.3/i386-linux- 
thread-multi/CORE/libperl.so
#22 0x006f46a5 in perl_run () from /usr/lib/perl5/5.8.3/i386-linux- 
thread-multi/CORE/libperl.so
#23 0x08049394 in main ()



The code from that point is all unfamiliar to me, and not immediately  
obvious.  gtk_entry_completion_insert_prefix() is attempting to block  
the insert_text signal on the completion's entry before emitting the  
insert-prefix signal, but the entry has no such signal handler  
connected.  The insert_text_id is captured in  
connect_completion_signals() in gtkentry.c when the completion is  
attached to the entry, and is disconnected in  
disconnect_completion_signals()... but the value doesn't appear to be  
cleared from the instance.  Not sure if it's a garbage or you were  
supposed to destroy the completion or what.

That's as far as i can take it at the moment... really need sleep...