GitLab

GnomeCanvas will stash the GdkDevice and reuse it in the subsequent
gnome_canvas_item_ungrab() call.

Prefer dealing with GdkEvent pointers and using accessor functions like
gdk_event_get_button().

This is complicated by the fact that some GtkWidget method declarations
still use GdkEventButton pointers, and synthesizing button events pretty
much requires direct GdkEventButton access.  But GDK seems to be nudging
itself toward sealing the GdkEvent union.  Likely to happen in GDK4.

Mainly clean up signal handlers and leave method overrides alone for now.

Adding temporary printf() calls to aid in debugging is fine, but leaving
them in the code permanently is not useful, especially when maintenance
duties are passed on to others.

Adding temporary printf() calls to aid in debugging is fine, but leaving
them in the code permanently is not useful, especially when maintenance
duties are passed on to others.

Will be needed for grabs in response to button events.

Use gdk_window_get_device_position() instead.

Use g_object_unref() instead.

Use gtk_scrollable_get_vadjustment() instead.

Use GLIB_VERSION_MAX_ALLOWED and GDK_VERSION_MAX_ALLOWED to warn if we
accidentally use symbols that were introduced in versions subsequent to
our minimum requirement.

Use GLIB_VERSION_MIN_REQUIRED and GDK_VERSION_MIN_REQUIRED to suppress
deprecation warnings by version so we can clean them up incrementally.
The initial values produce a few deprecation warnings which should not
be terribly difficult to clean up.

This fixes three things:
a) mail_display_resource_requested() was leaking file handles on downloaded images
b) it's not necessary to open files, it's enough to test whether file exists
c) mail_display_button_press_event() used incorrect key for "http" cache