Gtk+ 3

Gtk+ 3

[Andreas Enge]

Hello,

packaging of gtk+-3.10.0 is making progress. It compiles, and I think
only a few tests reading in /etc need to be disabled.

After that, to be in line with our general policy of trying to use the
latest versions of software, I think we should rename the variable gtk+
to gtk+-2, and call gtk+ the version 3.10.0. Then all packages requiring
gtk+ will be recompiled using 3.10.0; if this fails, we will need to adapt
the package inputs.

What do you think?

Andreas

Re: Gtk+ 3

[Ludovic Courtès]

Hello!

Andreas Enge <andreas@enge.fr> skribis:

> packaging of gtk+-3.10.0 is making progress. It compiles, and I think
> only a few tests reading in /etc need to be disabled.

Cool!  What kind of files are these?  For /etc/{services,protocols,rpc},
you could use net-base.

> After that, to be in line with our general policy of trying to use the
> latest versions of software, I think we should rename the variable gtk+
> to gtk+-2, and call gtk+ the version 3.10.0. Then all packages requiring
> gtk+ will be recompiled using 3.10.0; if this fails, we will need to adapt
> the package inputs.
>
> What do you think?

Sounds like a plan.  What about making the switch in a branch, and
adding a jobset on Hydra for that branch?  That way, if there’s too much
breakage, that allows things to be fixed calmly.

Thanks,
Ludo’.

Re: Gtk+ 3

[Andreas Enge]

On Tue, Oct 01, 2013 at 02:12:13PM +0200, Ludovic Courtès wrote:
> Cool!  What kind of files are these?  For /etc/{services,protocols,rpc},
> you could use net-base.

The error messages are as follows:
D-Bus library appears to be incorrectly set up; failed to read
machine uuid: Failed to open "/etc/machine-id": No such file or
directory. See the manual page for dbus-uuidgen to correct this issue.

> Sounds like a plan.  What about making the switch in a branch, and
> adding a jobset on Hydra for that branch?  That way, if there’s too much
> breakage, that allows things to be fixed calmly.

That would have made sense, but so far, we have only seven packages using
gtk+ (two if which, racket and xnee, turn out to also work with gtk+-3),
so I tested them by hand and just pushed the patch.

Andreas

Re: Gtk+ 3

[Ludovic Courtès]

Andreas Enge <andreas@enge.fr> skribis:

> On Tue, Oct 01, 2013 at 02:12:13PM +0200, Ludovic Courtès wrote:
>> Cool!  What kind of files are these?  For /etc/{services,protocols,rpc},
>> you could use net-base.
>
> The error messages are as follows:
> D-Bus library appears to be incorrectly set up; failed to read
> machine uuid: Failed to open "/etc/machine-id": No such file or
> directory. See the manual page for dbus-uuidgen to correct this issue.

Blech, what’s this file?  I don’t feel the need for another unique
identifier for my machine.  ;-)

Anyway, we probably can’t do much other than skipping these tests.

>> Sounds like a plan.  What about making the switch in a branch, and
>> adding a jobset on Hydra for that branch?  That way, if there’s too much
>> breakage, that allows things to be fixed calmly.
>
> That would have made sense, but so far, we have only seven packages using
> gtk+ (two if which, racket and xnee, turn out to also work with gtk+-3),
> so I tested them by hand and just pushed the patch.

OK, makes sense (somehow I thought we had plenty of GUI packages ;-)).

So that means GTK+ 3 actually works, right?  That’s cool!

Ludo’.

Re: Gtk+ 3

[Andreas Enge]

On Wed, Oct 02, 2013 at 12:14:47AM +0200, Ludovic Courtès wrote:
> So that means GTK+ 3 actually works, right?  That’s cool!

This is a good question, since none of the packages I am interested in
uses gtk+ so far. But I am working on packaging one ;-)

I tried to use the slideshow binary from the racket package; it fails with
ffi-lib: couldn't open "libgdk-x11-2.0.so.0" (libgdk-x11-2.0.so.0: cannot open shared object file: No such file or directory)
  context...:
So it also requires gtk+-2; apparently an error in the configure script.
I will also revert this back.

xnee has only one binary, cnee, which fails for me with
Error number:   6
  Error:        Wrong paramters
  Description:  You have given invalid input to Xnee 
  Argument:     (NO ARG)
At least, it does not look gtk+-related. On the other hand, "ldd cnee"
does not show anything gtk+-related either, nor does "guix gc --references";
so maybe the input gtk+ is not actually used at all?

So we should call for people to package gtk+-3 applications!

Andreas

Re: Gtk+ 3

[Ludovic Courtès]

Andreas Enge <andreas@enge.fr> skribis:

> On Wed, Oct 02, 2013 at 12:14:47AM +0200, Ludovic Courtès wrote:
>> So that means GTK+ 3 actually works, right?  That’s cool!
>
> This is a good question, since none of the packages I am interested in
> uses gtk+ so far. But I am working on packaging one ;-)
>
> I tried to use the slideshow binary from the racket package; it fails with
> ffi-lib: couldn't open "libgdk-x11-2.0.so.0" (libgdk-x11-2.0.so.0: cannot open shared object file: No such file or directory)
>   context...:
> So it also requires gtk+-2; apparently an error in the configure script.
> I will also revert this back.

OK.

> xnee has only one binary, cnee, which fails for me with
> Error number:   6
>   Error:        Wrong paramters
>   Description:  You have given invalid input to Xnee 
>   Argument:     (NO ARG)
> At least, it does not look gtk+-related. On the other hand, "ldd cnee"
> does not show anything gtk+-related either, nor does "guix gc --references";
> so maybe the input gtk+ is not actually used at all?

I think Xnee has an optional gnee program (or something like that) that
provides a GTK+ GUI.  But maybe it explicitly checks for gtk+-2.0 and
just disables that when that is not found?

I’ll check if Geeqie works and report back.

Ludo’.

Re: Gtk+ 3

[Andreas Enge]

On Wed, Oct 02, 2013 at 02:16:42PM +0200, Ludovic Court??s wrote:
> I’ll check if Geeqie works and report back.

Geeqie does not compile, whence commit 8b0275b6443.

Andreas

Re: Gtk+ 3

[Andreas Enge]

On Wed, Oct 02, 2013 at 02:16:42PM +0200, Ludovic Court??s wrote:
> I think Xnee has an optional gnee program (or something like that) that
> provides a GTK+ GUI.  But maybe it explicitly checks for gtk+-2.0 and
> just disables that when that is not found?

Even with gtk+-2, configure ends like this:

   Building the following components 
   -------------------------------------
    
        cli 
 
   Excluding the following components 
   -------------------------------------
    
        doc (docs are already included in dist file)
        gnee
        pnee 

The configure phase prints a line
   checking for gtk-config... no
Apparently, the program needs gtk1... See also
   http://savannah.gnu.org/bugs/?33666

I suggest to drop the input.

Andreas

Re: Gtk+ 3

[Ludovic Courtès]

Andreas Enge <andreas@enge.fr> skribis:

> On Wed, Oct 02, 2013 at 02:16:42PM +0200, Ludovic Court??s wrote:
>> I think Xnee has an optional gnee program (or something like that) that
>> provides a GTK+ GUI.  But maybe it explicitly checks for gtk+-2.0 and
>> just disables that when that is not found?
>
> Even with gtk+-2, configure ends like this:
>
>    Building the following components 
>    -------------------------------------
>     
>         cli 
>  
>    Excluding the following components 
>    -------------------------------------
>     
>         doc (docs are already included in dist file)
>         gnee
>         pnee 
>
> The configure phase prints a line
>    checking for gtk-config... no
> Apparently, the program needs gtk1... See also
>    http://savannah.gnu.org/bugs/?33666
>
> I suggest to drop the input.

Agreed.

Ludo’.