At Tue, 12 Jun 2012 23:56:10 +0800, Marguerite Su wrote:
On Tue, Jun 12, 2012 at 11:17 PM, Takashi Iwai <tiwai@suse.de> wrote:
At Tue, 12 Jun 2012 23:06:43 +0800, Marguerite Su wrote:
On Tue, Jun 12, 2012 at 10:55 PM, Takashi Iwai <tiwai@suse.de> wrote:
At Tue, 12 Jun 2012 00:28:40 +0800, Marguerite Su wrote:
On Tue, Jun 12, 2012 at 12:19 AM, Takashi Iwai <tiwai@suse.de> wrote:
At Mon, 11 Jun 2012 23:44:59 +0800, Marguerite Su wrote: > > >> Some of the GTK functions and macros used in them are too old (GTK > >> 2.12) and have no new alternatives. > >> > >> It means we have to rewrite some lines of its codes. like > >> GTK_Fileselection to GTK_Dialog, GTK_combo to GTK_combobox. > >> > >> So if anyone have time and GTK coding experience, please offer some help. > > > > Hmm, how other distros handle this update? Haven't they hit the same > > problems? I see both Ubuntu and Arch already updated to scim 1.4.13, > > but they can build scim-anthy as is. > > > > Judging from the compile errors, isn't it just a side-effect of using > > GTK3 in SCIM...? > > That's the most interesting part: > > I see Fedora and Ubuntu have 1.4.13 and all other packages the same > version as us. > > But they didn't patch anything. they just build fine. that part confuses me. > > I'm not that fresh to packaing and patching...those functions in them > are from gtk 2.12 era, which are already dropped for a long time. why > they can built it without any problem...
I guess other distros don't use GTK3 for SCIM but keep using GTK2. The incompatible part is likely the change of widget components in file-selection or combo, i.e. somewhat internal in GTK widgets.
maybe...
there're three kinds of errors:
- gtktooltips ( most easy)
- file-selection ( a little bit complicated but understandable)
- combo to combobox
what a mess...there're so many functions no longer available...I don't know how to deal with them, like:
gtk_combo_set_value_in_list gtk_combo_set_case_sensitive gtk_combo_set_popdown_strings
do you have any ideas?
https://build.opensuse.org/package/show?package=scim-skk&project=home%3A...
That's not trivial, unfortunately...
Looking at more on scim-chewing code, especially scim_color_button stuff is fairly hard to port to GTK3 as it's using lots of deprecated (and actaully dropped) functions. Definitely it isn't worth to waste time for doing that.
Yes...I tried to fix GtkObjectClass errors, then It comes a lot of other errors.
GtkObjectClass is a simple thing. It can be changed mostly to GtkWidget. And expose event can be changed to draw. But the actual rendering part, especially the implementation of the own color-selection button is harder to port. You'd need to convert gdk thingies there to cairo helpers, or equivalent ones.
Swyear asked taiwan developers to fix it. maybe I can wait for a while to hear his results.
So, maybe reverting to gtk2-devel in scim.rpm would have solved the breakage in scim-* subpackages automatically. But you've already fixed so many packages, and it's better to go foward at this point than taking back your previous works.
then maybe in the end we M17N team finish the work of porting scim related packages to gtk3...when I started, I didn't even know I was porting anything, I thought I was just fixing build errors...
Yeah, that was the fact. And, looking at how the GTK3 porting can be intrusive, desprite of my previous comment, I'd suggest to go back to GTK2 in scim packaging. The necessary change is simple, just pass --with-gtk-version=2 to configure script in scim.
SCIM is an old stuff, as you already mentioned. In such a case, there is no big reason to take a huge risk of regressions by untested updates.
I prepared scim updates now with GTK2 in OBS home:tiwai:branches:M17N repos. All builds look OK, so far.
I saw that on OBS index when you branched. maybe we can take your branch as a failsafe and see what my branch can do?
now I have scim-anthy, scim-pinyin, scim-hangul, scim-skk, scim-unikey, scim-canna, novell-pinyin fixed(or ported)...all the things left are scim-uim and scim-chewing. maybe if community and I work a little harder...dream will come true...LOL
Of course, works to fix these are really appreciated, but honestly speaking, I really don't think it's woth to spend more time on that. Better to improve fcitx, instead :)
Haha, to talk about fcitx, fcitx-anthy is nearly ready. (weng stayed up for about 3 to 4 nights and get it done)
it's now in an inside testing state.
maybe later this week I'll package it to M17N:Devel
OK.
GTK2 won't be dropped so soon, so we can keep using it. (Note that --with-gtk-version=2 option doesn't stop building GTK3 IM module of SCIM. It's just that scim itself (and library) is linked with gtk2 instead of the default gtk3.)
Yes...if you told me a little earlier...it'll save me a lot time...
Oh, if you told me the problme a bit earlier... ;)
but...now I even fixed novel-pinyin...only three to four left...it's really hard to give up...
Well, my concern is about the stability. I guess you've just fixed builds but not tested the actual system so much. Such a program is difficult to test actually if you don't know of the exotic operation. If the changes are only in SCIM, you can test one of the component you are familiar with (i.e. the Chinese IM) to see whether any regression happens. But for all these components? It's a lot of QA works. So, I guess that a safer way at this moment is a smaller change.
Actually except for scim-chewing, other fixes are just function migration like gtktooltips..not rewrite or something, so it won't affect stability too much...
Yes, I looked through your patches (at least the ones in M17N), and these look OK. The patches work even with the recent GTK2, except for the version in 11.4. So, I keep these in the packages in my branched ones. Takashi -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org