glib/GTK update tentative [Split from Pidgin thread]


glib could probably be updated in OS, it shouldn't be as problematic as updating something like QT.
OOOh ! That would be great if you could do that, as many things are rather difficult to compile because they need a more recent glib (like at least a 2.30).
 
glib could probably be updated in OS, it shouldn't be as problematic as updating something like QT.
OOOh ! That would be great if you could do that, as many things are rather difficult to compile because they need a more recent glib (like at least a 2.30).
And if you do so, update GTK too please.
 
Ok compiled glib 2.34, installed on my pandora and Xfce no longer starts, seems to be crashing.

glib is really not keeping binary compatibility, good thing I haven't pushed out an update for everyone..

About making different versions coexist, OE doesn't offer such functionality, and it's like creating a fork anyway, maintaining compatibility for these PNDs linked against a fork will only bring headaches in future. It won't solve the problem anyway - say some PND is linked against a forked glib and a few other libs in firmware; but one of those firmware libs is linked against "main" glib. You end up having both glibs loaded into the same process, lots of duplicate name functions (but with different content) in symbol table, and the dynamic linker will link random function to your program, so you crash anyway.
 
Last edited by a moderator:
Well, if you can figure it out, great! If not, oh well at least you tried. Right?
 
I don't really understand, why they don't release ABI changed versions with a new name like glibc234.so or similar... However: Static linking?
 
Last edited by a moderator:
Sort of a progress report from last night:

After recompiling xfce4-session Xfce started again, but it's broken all over the place, gnome-keyring errors out so no wifi, even terminal doesn't start. Logs show lots of errors from gtk, so I tried recompiling it but of course it failed because it relies on some macros in old glib. Then tried updating gtk, but problems there again as newer gtk was split into 2 libs (gdk-pixbuf split out). Recipe from newer OE failed but after more hacking it finally compiled, after more pain gtk compiled too. Starting OS build now resulted in yet another failure of atk lib, again because of glib, updating that we get glade failing to build, but haven't checked that one yet.

So it's horrible dependency hell, I wish I never started this. Makes me wonder how distribution maintainers cope with all this, I guess picking all newest versions of everything results in somewhat working system as the developers are using everything this way, but it seems updating single gnome component is just suicide.
 
Ah well, I wouldn't bother then. Will the Pyra at least have the right version?
 
Sort of a progress report from last night:


After recompiling xfce4-session Xfce started again, but it's broken all over the place, gnome-keyring errors out so no wifi, even terminal doesn't start. Logs show lots of errors from gtk, so I tried recompiling it but of course it failed because it relies on some macros in old glib. Then tried updating gtk, but problems there again as newer gtk was split into 2 libs (gdk-pixbuf split out). Recipe from newer OE failed but after more hacking it finally compiled, after more pain gtk compiled too. Starting OS build now resulted in yet another failure of atk lib, again because of glib, updating that we get glade failing to build, but haven't checked that one yet.


So it's horrible dependency hell, I wish I never started this. Makes me wonder how distribution maintainers cope with all this, I guess picking all newest versions of everything results in somewhat working system as the developers are using everything this way, but it seems updating single gnome component is just suicide.
I do agree and share the pain! I tried also to updated glib and friends for codeblocks... And yes, dependencies hell!
 
Only solution I could think of is to install them seperately in /usr/local for example and package them in a pnd that overlays them to the correct system path when mounted. So application maintainers need to get them from there if newer programs need them.
 
Only solution I could think of is to install them seperately in /usr/local for example and package them in a pnd that overlays them to the correct system path when mounted. So application maintainers need to get them from there if newer programs need them.
It still won't work if the PND loads any "usual" libs that use glib too, it all will just crash.
 
Only solution I could think of is to install them seperately in /usr/local for example and package them in a pnd that overlays them to the correct system path when mounted. So application maintainers need to get them from there if newer programs need them.
It still won't work if the PND loads any "usual" libs that use glib too, it all will just crash.
Then it just needs to include ALL libs that use glib inside the PND. Yes, that will make the PND very large, but it's worth it.
 
I built glic, gtk3 and all the dependencies in codeblocks and installed them in a separate folder. Then compiled apps against them and it was working fine actually.
 
Sort of a progress report from last night:


After recompiling xfce4-session Xfce started again, but it's broken all over the place, gnome-keyring errors out so no wifi, even terminal doesn't start. Logs show lots of errors from gtk, so I tried recompiling it but of course it failed because it relies on some macros in old glib. Then tried updating gtk, but problems there again as newer gtk was split into 2 libs (gdk-pixbuf split out). Recipe from newer OE failed but after more hacking it finally compiled, after more pain gtk compiled too. Starting OS build now resulted in yet another failure of atk lib, again because of glib, updating that we get glade failing to build, but haven't checked that one yet.


So it's horrible dependency hell, I wish I never started this. Makes me wonder how distribution maintainers cope with all this, I guess picking all newest versions of everything results in somewhat working system as the developers are using everything this way, but it seems updating single gnome component is just suicide.
In debian glibc is behind, and xfce4 isnt new. I hope LXqt can replace it so that one is more free to ship updated libraries.
 
glibc is not involved, it's doing a real good job maintaining the ABI and has versioned symbols in case they need to change anything.

It's GLib (GNOME Lib?).
 
Ok after lots of pain I have got the latest glib and gtk2 building, xfce seems to be booting and working.
 
I hope I can get some help with testing before it's all pushed to main OS update as lots of other libs had to be updated to build and work with latest glib.
 
To test, you have to change /etc/opkg/openpandora.org-feed.conf to point to my test feed instead:

Then run OS update as usual.
 
Edit: everything has been pushed to official feeds as 1.62-rc, see 1.62 thread:

http://boards.openpandora.org/topic/17176-superzaxxon-v162rc-released/
 
Last edited by a moderator:
Just upgraded as well and have to give out a small warning:

After upgrading and rebooting, XFCE4 didn't run (always went back to the login screen).

I booted into CLI mode and renamed my home-directory in order to get a clean one.

After doing that, it booted up find and is working without issues so far.

On my old profile, I changed some things such as font sizes, etc.

Seems like the profiles are not 100% compatible, so for an official update, we probably need to delete custom XFCE4-Settings before the upgrade.
 
Back
Top