Arch Linux

My locale is pt_BR.utf-8, and my physical keyboard layout is US. I use the "us-acentos" keymap in the console, and the corresponding us_intl in X. Pressing (dead) acute then "c" has produced "ç" (that's "c" with a diacritic called "cedilla" in Spanish or "cedilha" in Portuguese) reliably for me since the 90's, in every operational system and every distribution, but it stopped working after I updated GNOME to 3.6 with Arch Linux. Now the default behavior dead acute + "c" = "ć". Other people from my locale could reproduce this issue with Arch Linux, and people in my locale using other distributions are apparently still using GNOME 3.4. It affects other applications (e.g. Firefox, Skype) in a GNOME session, as well as typing the username in GDM.

I tried to clear my previous keyboard configuration (in GNOME) and redo it, either using the graphical user interface or with gsettings, but the result was the same. I thought it had something to do with ibus, but ironically selecting the ibus input method from the context menu in GTK+ applications makes the combination work as expected for me. Actually, the workaround was adding GTK_IM_METHOD=ibus to my (previously absent) ~/.profile.

Needing to edit a configuration file in order to get basic stuff working is simply not how GNOME should work. Either this is a regression in GNOME itself, or an issue in how it was packaged. I'm reporting this first to Arch Linux to ask for confirmation before submitting a bug report upstream.

This could probably be avoided with a GTK_IM_METHOD=ibus entry in /etc/profile.d/, but of course it's up to Arch Linux and GNOME/IBus to decide on such matters.

Comment: The "us-acentos" keymap file in the kbd package (my version: 1.15.3-3) defines ' + c = ç, the "us" symbols file in the xkeyboard-config package (my version: 2.7-1) defines as ' + c = ć, and the pt_BR.UTF-8 locale Compose file in the libX11 package (my version 1.5.0-1) defines ' + c = ç.