Slight issue with the previous commits. I've edited the *.pc.in files that generate the final *.pc file rather than dynamically creating them based on the results of the PKG_CHECK_MODULES result. osso-systemui is also affected as it hard codes both GTK and Hildon versions.

My somewhat ugly solution at the moment is to define a hildon_modules value as part of the succeed stage and then use AC_SUBST to replace a reference to @hildon_modules@ in the *.pc.in file. There is probably a far more elegant/clean solution but it works for GTK3, not tested on GTK1.

This is an example from libhildondesktop:
I may look to having a --with-gtk3 flag to configure as well. Otherwise it forces GTK3 if it and libhildon3 are installed when you might be trying to build for GTK2/libhildon1.

clockd is finished - https://github.com/community-ssu/clockd

almost not tested, but it shouldn't matter much at that stage I guess

@Android_808: What is the state on your side? Does something else need to be REed?

I've been reading up a bit this week, trying to make head to tail of some of the other components.

To start with, I've been looking into I started to look at hildom-input-method-framework (from http://mirrors.muarf.org, not sure if original repo still exists). I've done quite a bit of GTK3 porting using a WITH_GTK3 like in osso-systemui but I've a few issues.

First, some code attempts to call routines either hidden behind MAEMO_CHANGES or are expected to be in a modified (non-Maemo) hildon version of gtk2. I can edit it but don't know what the results will be.

Second I haven't got a clue how it works/ what its doing yet. The im_context parameters it's trying to mess with in GtkEntry etc. are private, but they may have been exposed by Maemo GTK patch ,and most of the methods using them like gtk_im_context_hide were depreciated according to hildon's 2.6 notes. I don't know whether I just need to adapt it to assign modules by name, scrap it or what.

I might start looking at updating some of the init scripts or looking at possibly implementing optional systemd support in dsme based of Mer version. The GTK3/configure.ac/HAVE_GTK3 set up is another aim moving forward.

RE wise, I'm not 100% sure at the moment. We are missing a few of the osso-systemui-* modules (splashscreen and devlock) . We do have headers in CSSU already but not the binaries. Might be worth looking at the closed package list (there is a google cache version if needed due to site maintenance) to see if there is anything obvious.

On a sad note today, trying to look up prices online whilst out has really frustrated me on the N900. Mainly because the slow response and incorrect web rendering wouldn't let me log in, resulting in missing out on a bargain on Amazon. I even popped into one of the stores and took a look at a OnePlus 3 (3T would be ordered) and a Pixel/Pixel XL. None support microsd, and the 32Gb Pixel only has 23Gb available. I have 288Gb (minus Maemo/Apps/Swap) to play with at the moment if I want, although I'm only running 96Gb. Unlike the N95 and N900 there's nothing screaming "buy me", nothing that ticks all my requirements.

Do we really need hildon-im-context right now? Is everything else hildon ready? As I just tried to build hildon-desktop from your repo and failed miserably . Maybe you did not push everything.

I know it doen't have the hildon-1 to hildon-3 change done yet. just change it in configure.ac.

With the current control file it needs building with dpkg-buildpackage -rfakeroot -us -uc -b -d. I have't finished maemo-launcher yet, it is disabled in h-d rules but not in control.

Give that a try and if it fails let me know the error.

edit: the mer hack! might not have reverted/edited changes there yet. it is a single commit. one/two mce/dsme lines changed the rest commented. i'll look some stage today.

Which clutter shall I use? what about libmatchbox2?


EDIT: BTW, could you /join #maemo on freenode?

latest clutter (1.2x iirc) and libmx from debian jessie, haven't tried stretch yet. libmatchbox2 from my repo (has gtk3 fixes).

might be going out in a bit so i'll be unavailable till later unless i get irc set up on phone

first results, devuan jessie on allwinner tablet (mali 400 GPU)

http://46.249.74.23/allwinner/20161127_001.mp4

(download the video, do not try to play it in the browser)

@Android_808: after some play on the tablet, so far:

"touch-event" needs to be used along with "button-release-event" etc, otherwise it does not work on touchscreen.

I don't know if cordia devs broke it or it is because of migration to clutter 1x, but task navigator barely works, if at all. And as tasknav code is like spaghetti anyways, I think the viable option is to just scrap cordia h-d and reapply the needed changes on top of cssu h-d, instead of debugging what is broken. Not to say there are #if MAEGO_DISABLED chunks all over the place, mce/dsme code is removed, call-ui code is removed and whatnot.

What do you think?

I didn't hit the touch-event issue as I was just using virtualbox. I know there is a rendering issue with bottom of each preview transparent and a race condition if you switch between app/switcher too quickly causing window to disappear.

I don't mind either way, the task switcher is going to be a real pain either way because the original is strung together using code that is now deprecated..at least twice. Hence the current mess. LIbhildon would need to use CSSU version as well, and would probably have to be done first. I'm still not 100% sure on solution for hildon-fm/filechooser.

My original aim with the project was to see what a future direction could be, so stuff like total backwards compatibility was not a major priority. For some features it just isn't worth it, like the touch support. Just check HildonPannableArea. Maybe going through and assessing each of the deprecated features in the original code and making a decision.

If using CSSU release, would it be worth ditching mx as well and using an updated version of tidy, similar to what champlain does (https://github.com/GNOME/libchamplai...ster/champlain)

On a side note, I did look at updating upstart-dev to install to the correct location. One change is it just needs event.replace.d and event.d replaced with init.replace.d and init. The second is to append .conf to the filename, but for some reason that didn't want to work.

I'm back at work this week so won't be around during the day.