GtkSelectionModel Interface not always correct

[jeremynesbitt]

Describe the bug
I was making a test program to evaluate the GtkColumnView implementation and found that the fortran interface to
gtk_column_view_new(selection) is incorrect
Expected behavior

In gtk-auto.in, should have the following. I think there are other instances where GtkSelectionModel has a c_int interface instead of c_ptr. I tried to find the logic in your python scripts to change it upstream, but I was unable to figure out how they worked well enough to quickly fix it.
! GDK_AVAILABLE_IN_ALL
!GtkWidget * gtk_column_view_new (GtkSelectionModel *model);
function gtk_column_view_new(model) bind(c)
import :: c_ptr, c_int
implicit none
type(c_ptr) :: gtk_column_view_new
!integer(c_int), value :: model this is the incorrect interface
type(c_ptr), value :: model
end function
To Reproduce
Steps to reproduce the behavior:
I have attached a sample program (I think), which is the gtkhello.f90 example with a columnview object grafted on.
If interested, you can run this by extracting to a folder and using meson to build.
mkdir build
cd build
meson setup ..
meson compile
`
It will not compile without making the change I specify above. After making the change, the object is successfully created and displays the information as expected.

gtk4ColumnViewExample.zip

Your system:
gtk-4-fortran version 4.7

Additional context
Add any other context about the problem here.

[vmagnin]

Thanks @jnez137 for reporting this problem, and for the example.

I will investigate in the next weeks and try to understand why the cfwrapper is not considering model as a C pointer although there is a star *.

[vmagnin]

• https://docs.gtk.org/gtk4/ctor.ColumnView.new.html
• https://docs.gtk.org/gtk4/iface.SelectionModel.html

[vmagnin]

The problem is the same in all functions having an argument GtkSelectionModel *model.

When there are several arguments with a star, the others are treated correctly:

!void gtk_grid_view_set_model (GtkGridView *self, GtkSelectionModel *model);
subroutine gtk_grid_view_set_model(self, model) bind(c)
  import :: c_ptr, c_int
  implicit none
  type(c_ptr), value :: self
  integer(c_int), value :: model
end subroutine

self is correctly declared, not model.

Same here, factory is OK, not model:

!GtkWidget * gtk_grid_view_new (GtkSelectionModel *model, GtkListItemFactory *factory);
function gtk_grid_view_new(model, factory) bind(c)
  import :: c_ptr, c_int
  implicit none
  type(c_ptr) :: gtk_grid_view_new
  integer(c_int), value :: model
  type(c_ptr), value :: factory
end function

Header file: /usr/include/gtk-4.0/gtk/gtkselectionmodel.h

[vmagnin]

In /usr/include/gtk-4.0/gtk/gtkgridview.h:

GDK_AVAILABLE_IN_ALL
GtkWidget *     gtk_grid_view_new                               (GtkSelectionModel      *model,
                                                                 GtkListItemFactory     *factory);

GDK_AVAILABLE_IN_ALL
void            gtk_grid_view_set_model                         (GtkGridView            *self,
                                                                 GtkSelectionModel      *model);

[vmagnin]

After investigation, I think the problem is that there is in src/gtkenums-auto.in an enum with nearly the same name (without an l at the end):

enum, bind(c)    !GtkSelectionMode
  enumerator :: GTK_SELECTION_NONE
  enumerator :: GTK_SELECTION_SINGLE
  enumerator :: GTK_SELECTION_BROWSE
  enumerator :: GTK_SELECTION_MULTIPLE
end enum

In def iso_c_binding(declaration, isReturned): in cfwrapper/fortran.py, I will have to rework probably those lines:

    # Is it a "typedef enum" ?
    for item in types_enums.gtk_enums:
        if item in c_type:
            return "integer(c_int)", "c_int"

The if item in c_type: is the problem.

But maybe, moving this block after the block treating the * could be sufficient. I need to think and test the side effects. Well, that's a good progress for today.

[vmagnin]

Thanks @jnez137, the interfaces of 98 functions are now fixed (~1% of the library). In cfwrapper/fortran.py, the block treating the enums was not placed at the end of the iso_c_binding() function: first, C pointers toward enums were not recognized as pointers, and secondly enums having a name included in a GTK type name could cause bad bindings (for example GtkSelectionMode and GtkSelectionModel).

I have pushed the new bindings in my gtk4-vmagnin branch: 25711aa

You can test it.
I will make a 4.7.1 release after some testing.

[jeremynesbitt]

Thanks Vincent! I'm happy to hear the library is 1% better now :) I checked out the branch and verified that the interfaces I have been using are working as expected.

…

On Thu, Feb 27, 2025 at 12:20 PM Vincent Magnin ***@***.***> wrote: Thanks @jnez137 <https://github.com/jnez137>, the interfaces of 98 functions are now fixed (~1% of the library). In cfwrapper/fortran.py, the block treating the enums was not placed at the end of the iso_c_binding() function: first, C pointers toward enums were not recognized as pointers, and secondly enums having a name included in a GTK type name could cause bad bindings (for example GtkSelectionMode and GtkSelectionModel). I have pushed the new bindings in my gtk4-vmagnin branch: 25711aa <25711aa> You can test it. I will make a 4.7.1 release after some testing. — Reply to this email directly, view it on GitHub <#290 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEVX64HDCVFGDL5WKHXNKO32R5XRJAVCNFSM6AAAAABXFCMPE6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMOBZGAZDQNRUHE> . You are receiving this because you were mentioned.Message ID: ***@***.***> [image: vmagnin]*vmagnin* left a comment (vmagnin/gtk-fortran#290) <#290 (comment)> Thanks @jnez137 <https://github.com/jnez137>, the interfaces of 98 functions are now fixed (~1% of the library). In cfwrapper/fortran.py, the block treating the enums was not placed at the end of the iso_c_binding() function: first, C pointers toward enums were not recognized as pointers, and secondly enums having a name included in a GTK type name could cause bad bindings (for example GtkSelectionMode and GtkSelectionModel). I have pushed the new bindings in my gtk4-vmagnin branch: 25711aa <25711aa> You can test it. I will make a 4.7.1 release after some testing. — Reply to this email directly, view it on GitHub <#290 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEVX64HDCVFGDL5WKHXNKO32R5XRJAVCNFSM6AAAAABXFCMPE6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMOBZGAZDQNRUHE> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[vmagnin]

I have merged the changes into the gtk4 branch and made a gtk-fortran 4.7.1 release.

@jnez137 , if you had already installed my dev branch, there is no need to update your installation. Nothing more than what you tested (thanks).