TODO.gtk2

5) Single click doesn't work in the fileselector (naturally use filechooser instead);
6) Huge CPU overhead (partially fixed);
7) Is very unstable (partially fixed). File->exit works good, but closing the 
main window leads to the crash;
9) Contains a lot of obsolete code.
10) Minor bug with selections, up and down buttons in the tracker fonts dialog
12) MIDI works something strange (also in 0.6.8-gtk1). It doesn't work at some systems
where it has to;
14) some CPU-consuming code in the instrument list selection callback (really 
in the callback);
15) Change option-menu in keys.c with notebook (like in audioconfig.c) for better
widget sizes' behaviour;
16) Fix mouse click at oscilloscopes (not only channel numbers) turn off/on channels;
17) Understand why tracker scrolling is such CPU-consuming (compared to ST-Gtk+-1). Maybe it's
a GTK+2.6.7 and is fixed in the modern release?;
19) Doesn't remember which font is selected for the tracker. To fix;
21) Native Alsa 1.x driver: make Alsa 1.x the requirement instead of 0.9.x; make sound device
opening non-blocking, but give error dialog instead when the device is busy; optimize GUI
(maybe with table) -- no idea :-(; problems with time estimation; signedness handling;
add avalable devices navigation (from csound->rtalsa.c);
22) ESD driver reported to produce greater delay at start than in ST-gtk1;
24) Chechk whether some str functions needed to be replaced with gtk_utf_(strlen,
offset_to_pointer and so on);
25) Return old-style tracker separation bars up to the top of the tracker (near tracker.c->660);
26) ESC for Cancel, Enter for Ok keybindings in dialogs;
28) Check and clean all gtk_dialog()s;
29) Provide more info in gtk_about_dialog

Regarding gtk_entry_set_text double signal: see gtkentry.c->2059, 2300 (workaround in the ST
sources is made. But a) we need to test whether it's fixed in Gtk+-2.10 and b) we should
push the fixes into the Gtk+-2.0 upstream. NEWS: it's at least fixed at the GTK+ upstream :)
So we need to know from what release this will happen and make corresponding check in
configure.in.

Some remarks about asm mixer bug: at the sample end and loop points there are ugly clicks
(ramping works badly). Also when the previous channel has amplitude > 0, the following one
will be triggered to the minimal negative value (maybe it's a cause of the first issue
'cause really ramping and looping is performing using additional channels. Probably the bug
is in the mixing or post-mixing code.