Mail Archives: geda-user/2015/10/03/22:48:50
On Sun, 4 Oct 2015, Evan Foss (evanfoss AT gmail DOT com) [via geda-user AT delorie DOT com] wrote:
<snip>
> Igor2 is off working on something using SDL because it is maintained
> but has a slower release cycle. The logical reason being that we don't
> really care about new gtk functionality but we have to burn time
> fixing things for each major gtk release.
SPOLIER ALERT: I am not talking about gschem or pcb here; I don't say any
of the geda tools should or should not use gtk. I just share my personal
opinion about using gtk in general. I won't respond to text that ignores
this.
Yes, I think GTK is a bad idea long term (for any application), and it's
not qt that'd solve this. My resoning goes:
- When gtk changes API, sooner or later projects using gtk have to change
too, even if there's no real reason for the change; this often involves
not only technical code change, but gui redesign on the app side. To me
it seems like a waste to do major changes on the GUI not because the app
needs it but gtk needs it.
- I know I am not a typical user, but I hate eye-candy changes, and
majority of such a gtk change is usually about that. So this is changing
the GUI of an applicaiton just that it looks more modern while I'm still
trying to do the same job as 10 years ago. I don't say the GUI or any
other part should be frozen just because I got used to it, but I do say
the desire for the change should be coming from the application not
from a GUI toolkit change.
- GTK doesn't look native on anything else than gnome. Most other toolkits
have the same property: native only on one or on a few systems. So using
anything else doesn't make this worse or better. (If an app wants to look
native on multiple platforms, it needs multiple native GUI frontends, one
for each platform.)
- GTK is huge: gtk+2.0 is over 368k sloc. For the comparison, pcb-rnd
trunk/ is about 160k and geda-gaf is below 120k. So having buttons and
menus costs more code than the two tools combined - not counting the
dependencies of GTK.
- Future of GTK doesn't look good either: gtk+3.0 is 672k sloc. I
don't believe version 4 will be smaller than version 2, even if it'd be
about cleanup.
- GTK brings other stuff like glib. I know my opinion is being extreme
again, but I don't think glib is a good idea in its current form. I also
believe it did nothing good for pcb or geda. Debian says gtk 2.0
also depends on libs like glib, gdk, pango, atk, cairo.
- Portability; GTK and other major toolkits are usually ported to a few
popular systems by the maintainers so projects don't need to worry about
that, which is a good point. But if you have a system it is not ported to,
you'll probably have hard time porting it. Especially if you need to also
port the dependencies... Does GTK work well on tablets?
For my own amusement I'm working on a small UI toolkit based on SDL2. The
SDL project doesn't have major API changes as often as gtk, and is much
smaller. It is ported to at least as many systems as gtk is - inlcuding
Linux, modern BSD systems (e.g. osx), windows, android and iOS. Looking at
video/, porting to a new system doesn't seem to be very hard either. SDL
doesn't bring excess dependencies.
With the SDL based toolkit, eventually I will have an SDL based HID
for pcb-rnd. I believe it will be faster than the gtk hid, it will be more
customizable (GUI layout will be data, not code), and more portable
(windows cross compiles will be much simpler). It'll take a long time to
finish the toolkit, tho.
Regards,
Igor2
- Raw text -