[WIP] Fremantle GTK3 Experiment - Page 4 - maemo.org

Active Topics

Sailfish OS for the Motorola Moto G7 Power (XT1955-5) - (ocean) (3)
to SailfishOS by edp17 - 4 hrs, 49 mins ago
The N900 is now 15 years old! (6)
to Nokia N900 by Macros - 1 day, 8 hrs ago
F(x)tec Pro1x (QWERTY) for sale (0)
to Buy & Sell by aln00000 - 1 day, 21 hrs ago
n770 youtube video (0)
to General by nmdv - 5 days, 20 hrs ago
more...

Page 4 of 19

Thread Tools

Android_808	2016-11-12 , 08:17
Posts: 1,203 \| Thanked: 3,027 times \| Joined on Dec 2010	#31

thanks. didn't have much time to look into it yesterday.

I think i just need to rebuild dsme and libdsme, then build osso-systemui, libconic and then i can replace mce. then i just need to revert the hildon-desktop mer hack commit. i might have a play with dsme and mce to see if i can add optional support for systemd as mentioned earlier.

i have also freed up some space on the laptop for another build. as my working version is built on amd64 jessie, i'm going to set up x86 stretch to test for issues going forward and give me a way to work on ports when away from home.

Quote & Reply |

The Following 6 Users Say Thank You to Android_808 For This Useful Post:
DA5, freemangordon, juiceme, peterleinchen, wicket, Wikiwide

freemangordon	2016-11-12 , 15:38
Posts: 3,074 \| Thanked: 12,960 times \| Joined on Mar 2010 @ Sofia,Bulgaria	#32

iphbd is fully REed and mostly tested, make sure to pull before playing further as there were some fixes in libiphbd0

https://github.com/community-ssu/iphbd

__________________
Never fear. I is here.

720p video support on N900,SmartReflex on N900,Keyboard and mouse support on N900
Nothing is impossible - Stable thumb2 on n900

Community SSU developer
kernel-power developer and maintainer

Quote & Reply |

The Following 8 Users Say Thank You to freemangordon For This Useful Post:
Android_808, Boemien, DA5, juiceme, peterleinchen, taixzo, wicket, Wikiwide

Android_808	2016-11-12 , 18:03
Posts: 1,203 \| Thanked: 3,027 times \| Joined on Dec 2010	#33

iphbd built fine.

Did you have any issues with libdsme-0.60.30+0m5?

I'm getting the following error message. It's an autogenerated file. The line in question in the .depends file is "-e ", the first of two. It almost looks like it is writing the param from the echo statement from lines 232-246 from Rules.make.

make[1]: Entering directory '/home/user/Projects/Maemo/libdsme-0.60.30+0m5'
.depend:156: *** missing separator (did you mean TAB instead of 8 spaces?). Stop.
/home/user/Projects/Maemo/libdsme-0.60.30+0m5/Rules.make:130: recipe for target 'all' failed

.depends content, lines 154-160(EOF):

Code:

 /usr/include/x86_64-linux-gnu/bits/waitstatus.h /usr/include/alloca.h \
 /usr/include/x86_64-linux-gnu/bits/stdlib-float.h
-e 
libdsme.a: protocol.o message.o
-e 
libdsme.so: protocol.o message.o 
libdsme_dbus_if.so: modules/dsme_dbus_if.o

Quote & Reply |

The Following 2 Users Say Thank You to Android_808 For This Useful Post:
juiceme, Wikiwide

freemangordon	2016-11-12 , 18:14
Posts: 3,074 \| Thanked: 12,960 times \| Joined on Mar 2010 @ Sofia,Bulgaria	#34

Originally Posted by Android_808

iphbd built fine.

Did you have any issues with libdsme-0.60.30+0m5?

I'm getting the following error message. It's an autogenerated file. The line in question in the .depends file is "-e ", the first of two. It almost looks like it is writing the param from the echo statement from lines 232-246 from Rules.make.

.depends content, lines 154-160(EOF):
Code:
 /usr/include/x86_64-linux-gnu/bits/waitstatus.h /usr/include/alloca.h \
 /usr/include/x86_64-linux-gnu/bits/stdlib-float.h
-e 
libdsme.a: protocol.o message.o
-e 
libdsme.so: protocol.o message.o 
libdsme_dbus_if.so: modules/dsme_dbus_if.o

yes, dash echo doesn't understand -e echo parameter

.

I had to edit rules.make file and remove "-e" from all the echos

EDIT:
it seems we can workaround that by using /bin/echo, something like

if [-x /bin/echo]
ECHO=/bin/echo
else
ECHO=echo
fi

(I am not very good at scripts, but I guess you get the idea)

Last edited by freemangordon; 2016-11-12 at 18:20.

Quote & Reply |

The Following 5 Users Say Thank You to freemangordon For This Useful Post:
Halftux, juiceme, peterleinchen, t-b, Wikiwide

Android_808	2016-11-12 , 23:46
Posts: 1,203 \| Thanked: 3,027 times \| Joined on Dec 2010	#35

yeah. I tried it that way earlier with some success. A lot of consensus is to replace echo with printf. My initial attempt to do this though caused the .a and .so sections to have no contents after the library name.

Funny thing is it seems to work ok on 1 of the 3 lines. I'll play with this a bit in the future but for now just want to get compiled and running. x86 stretch image is ready to start building packages so i'll try to pull some repos in the morning to build whilst out tomorrow.

Quote & Reply |

The Following 5 Users Say Thank You to Android_808 For This Useful Post:
Brodyga, juiceme, peterleinchen, t-b, Wikiwide

Android_808	2016-11-13 , 19:12
Posts: 1,203 \| Thanked: 3,027 times \| Joined on Dec 2010	#36

Apparently further from maemo mce support that I thought.

osso-systemui failing at the moment. The Cordia based libhildon has removed some HildonBanner related code for hildon_banner_show_information and hildon_banner_set_timeout. The later should still be achievable via setting property, but the first requires code to be reinstated or changed.

Several other packages (osso-systemui-powerkeymenu, osso-systemui-tklock to name but 2) require GTK3 port.

Every time you think your getting closer you find more packages, depending on more packages.

osso-systemui-tklock requires libalarm-dev (have sources) which requires libtime-dev (libtime0 is AFAIK another blob).

Finished setting up Stretch container this morning and pulled in a few repos to start building whilst I was out. It's not as straight forward building packages as it is for Jessie, but that has been my target environment for a while.

Maemo/Stretch Related Issues:
General:
gtkdoc-mktmpl (or similar) deprecated. Used in libhildon and libhildondesktop.

Packages:
icd2-osso-ic-dev: change compat 4>5

libcal: already reported, uncomment dh_mkshlibs or whatever its called in rules.

libconic: #if !GLIB_CHECK_VERSION(2,36,0) g_type_init(); #endif needed in test-*.c
change compat 4>5

libdsme0.2.0: change compat 4>5

libhildondesktop: gdk_window_set_background_rgba deprecated. Add CFLAGS+= -Wno-deprecated-declarations. Function removed from GTK as of ~16th Oct 2016

libosso: 4>5

libosso1-dbg: missing /usr/lib/debug/*.so*. Temporarily renamed libosso1-dbg.install

mce-dev: change compat 4>5

osso-systemui-dbus-dev: compat 4>5

profiled: change compat 4>5

Quote & Reply |

The Following 11 Users Say Thank You to Android_808 For This Useful Post:
Brodyga, endsormeans, freemangordon, Halftux, juiceme, misiak, peterleinchen, t-b, trx, wicket, Wikiwide

freemangordon	2016-11-14 , 06:54
Posts: 3,074 \| Thanked: 12,960 times \| Joined on Mar 2010 @ Sofia,Bulgaria	#37

ok, going to RE clockd/libtime0

Quote & Reply |

The Following 11 Users Say Thank You to freemangordon For This Useful Post:
Android_808, DA5, Halftux, handaxe, juiceme, peterleinchen, t-b, taixzo, trx, wicket, Wikiwide

Android_808	2016-11-14 , 09:01
Posts: 1,203 \| Thanked: 3,027 times \| Joined on Dec 2010	#38

Originally Posted by Android_808

osso-systemui failing at the moment. The Cordia based libhildon has removed some HildonBanner related code for hildon_banner_show_information and hildon_banner_set_timeout.

My bad, it does exist. osso-systemui's dbus.c has WITH_HILDON include before including config.h. From what I can tell, the configure sequence is detecting hildon so is not running the check for gtk3 the check for it will only be run if hildon is not found.

This isn't something I'm confident on dealing with but I think it should be like this:

Code:

PKG_CHECK_MODULES([HILDON], [hildon-1, gtk+-3.0],
    [AC_DEFINE(WITH_HILDON,[1],[Use Hildon]) AC_DEFINE(WITH_GTK3,[1],[Use Gtk3])],
    [PKG_CHECK_MODULES(HILDON, [hildon-1], [AC_DEFINE(WITH_HILDON,[1],[Use Hildon])])])

That gets it a bit further, but we then hit libcanberra DSO errors. $(CANBERRA_CFLAGS) and $(CANBERRA_LIBS) are missing from src/Makefile.am

Finally, -ldl appears to be missing from the same Makefile.am.

I don't know if the lib changes are exclusive to GTK3 or not at the moment.

Last edited by Android_808; 2016-11-14 at 09:14.

Quote & Reply |

The Following 5 Users Say Thank You to Android_808 For This Useful Post:
freemangordon, juiceme, peterleinchen, trx, Wikiwide

freemangordon	2016-11-14 , 11:46
Posts: 3,074 \| Thanked: 12,960 times \| Joined on Mar 2010 @ Sofia,Bulgaria	#39

Originally Posted by Android_808

My bad, it does exist. osso-systemui's dbus.c has WITH_HILDON include before including config.h.

Oh, I see, my bad, sorry about it.

From what I can tell, the configure sequence is detecting hildon so is not running the check for gtk3 the check for it will only be run if hildon is not found.

Right, this should be fixed.

This isn't something I'm confident on dealing with but I think it should be like this:

Code:

PKG_CHECK_MODULES([HILDON], [hildon-1, gtk+-3.0],
    [AC_DEFINE(WITH_HILDON,[1],[Use Hildon]) AC_DEFINE(WITH_GTK3,[1],[Use Gtk3])],
    [PKG_CHECK_MODULES(HILDON, [hildon-1], [AC_DEFINE(WITH_HILDON,[1],[Use Hildon])])])

Please clone systemui repo on github, do the required modifications and request back a merge. I am first to confess I don't feel comfortable with automake spaghetti.

That gets it a bit further, but we then hit libcanberra DSO errors. $(CANBERRA_CFLAGS) and $(CANBERRA_LIBS) are missing from src/Makefile.am

Finally, -ldl appears to be missing from the same Makefile.am.

I don't know if the lib changes are exclusive to GTK3 or not at the moment.

I'd like to keep one and the same code for both CSSU and hildon-gtk3 as much as possible - it really doesn't make sense to make forks for hildon-gtk3 when a couple of #ifdefs will do the job. Esp if we have a cooperative package maintainer to deal with (myself)

Quote & Reply |

The Following 5 Users Say Thank You to freemangordon For This Useful Post:
juiceme, mr_pingu, peterleinchen, trx, Wikiwide

Android_808	2016-11-14 , 13:22
Posts: 1,203 \| Thanked: 3,027 times \| Joined on Dec 2010	#40

I'd like to keep one and the same code for both CSSU and hildon-gtk3 as much as possible - it really doesn't make sense to make forks for hildon-gtk3 when a couple of #ifdefs will do the job. Esp if we have a cooperative package maintainer to deal with (myself)

I was thinking about this earlier. How do we manage the gtk2-dev/gtk3-dev dependency in the control file? The trouble with that is you could then end up trying to build with gtk2 headers and gtk3 hildon. Not sure if Debian control has a & option. If not it might be a case of dynamically generation the control file.

It will get a bit messy when it comes to libhildon and the rest of them because of the size of the code base compared to osso-* libraries.

I am first to confess I don't feel comfortable with automake spaghetti.

The automake spaghetti is above me. I just took a guess based on a few quick google results. For the missing libraries, would it be useful to try using pkg-config to pull them in as needed?

With regards to the Jessie port, osso-systemui is built. I had to go back to osso-af-utils and make a few hacks to anim-shower to bypass some outdated code. I'm not sure where/if the executable is used so it might come back to haunt me later.

Finally for today, upstart-dev. It doesn't exist in Debian, we have libupstart-dev. The Maemo version has dh_installxsession in it which is needed to finish building the rest of the osso-* pieces before I can install them. I need to look into how best to go about this, whether to just package the needed files

I have to be totally honest, using Mer userland seems sooo much easier

Quote & Reply |

The Following 8 Users Say Thank You to Android_808 For This Useful Post:
endsormeans, freemangordon, juiceme, mr_pingu, peterleinchen, taixzo, trx, Wikiwide