GitLab

Remove duplicate code and use a common function to set the mapping
between a numeric value and a text one. I think we should not
encourage the use of arbitrary text but instead allow the most common
ones. Therefore, add also 'off'.

If the VPN connection gets deactivated when the dialog is displayed,
the SecretsRequest is no longer valid. Accessing it in the dialog
callback causes a segmentation fault.

To fix this, disconnect the dialog handler when the secret request is
cancelled and install a new handler that just destroys the dialog.

https://bugzilla.redhat.com/show_bug.cgi?id=1775278

g_spawn_close_pid() is not necessary on Linux.

It is not necessary to check the validity of pid and then send the
signal; just send the signal.

Remove unnecessary comments.

These macros belong to libnma and no longer to applet.

This probably should be fixed by somebody giving love to
"page-mobile.c". For now, just ignore the deprecation warning.

It's a false positive. Avoid it.

  ../src/applet-dialogs.c: In function applet_info_dialog_show:
  ../src/applet-dialogs.c:709:21: warning: ip6_config may be used uninitialized in this function [-Wmaybe-uninitialized]
    709 |  dns6 = def6_addr ? nm_ip_config_get_nameservers (ip6_config) : NULL;
        |                     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  ../src/applet-dialogs.c:481:14: note: ip6_config was declared here
    481 |  NMIPConfig *ip6_config;
        |              ^~~~~~~~~~

Don't pass a floating variant to the callback function but convert it
into a full one so that we can guarantee that the callback function
will not assume ownership.

  (nm-applet:691510): GLib-CRITICAL **: 18:02:29.581: g_atomic_ref_count_dec: assertion 'g_atomic_int_get (arc) > 0' failed

  _g_log_abort () at /lib64/libglib-2.0.so.0
  g_logv () at /lib64/libglib-2.0.so.0
  g_log () at /lib64/libglib-2.0.so.0
  g_atomic_ref_count_dec () at /lib64/libglib-2.0.so.0
  g_variant_unref () at /lib64/libglib-2.0.so.0
  gs_local_variant_unref (v=<synthetic pointer>) at ./shared/nm-utils/gsystem-local-alloc.h:73
  complete_request (info=info@entry=0xaaf340) at src/applet-vpn-request.c:273
  external_ui_from_child_response (error=0x7fffffffd380, info=<optimized out>) at src/applet-vpn-request.c:261
  process_child_response (info=<optimized out>) at src/applet-vpn-request.c:298
  g_child_watch_dispatch () at /lib64/libglib-2.0.so.0
  g_main_context_dispatch () at /lib64/libglib-2.0.so.0
  ...

!77

[thaller@redhat.com: cherry-picked this patch from a larger
  merge request and trivial conflict resolution]

!73

Fix ce-ppp-auth-methods.ui as well per Mikhail's comment

!75 (comment 685501)

Fixes: 5fbf69f2 ('glade: get rid of GtkContainer.border-width')

This fixes a transparent border issue that appears to have been introduced
with commit 5fbf69f2:

With commit 5fbf69f2, ipv4/ipv6 route and new connection dialog boxes
were modified to use margin properties due to the deprecation of border
properties. Margin properties were placed under GtkDialog instead of the
immediate child GtkBox. Margin properties were also placed under another
GtkBox further below. This seemed to have the effect of margins not
being recognized on the top or left of the dialogs, but also creating a
transparent margin on the bottom and right. This was visible on the
ipv4/ipv6 route dialogs, but somehow also affected the new connection
dialog despite that file not having margin properties listed under
GtkDialog. Fixing properties in ce-ip4-routes.ui and ce-ip6-routes.ui
also fixed the new connection dialog.

Fixes: 5fbf69f2 ('glade: get rid of GtkContainer.border-width')

It looks more C-like and does does not compile code conditionally (while
the compiler still sees when compiled without appindicator support and
can strip the unreachable code).

This converts menu items from GdkPixbuf to a correctly scaled
cairo_surface_t so that it can render properly on HiDPI displays. This
change only applies to the GtkStatusIcon version of nm-applet, since
appindicator icon-data is currently set to handle GdkPixbuf.

Fixes: dc9ee4e1 ('applet: scale icons for HiDPI displays')

!74

!71

This partly reverts commit 94518cf9.

It's about time.

!68

This is used by WPA3 Personal and secured Meshes (which we don't support
in the applet).

#67
!63

The time stamp from the X server is not better than GDK_CURRENT_TIME
anyway.

#64
!64

Previously the secondary field would be focused even if the first was
empty.

When a connection is opened for editing in nm-c-e, secrets are asked
to NM. NM in turn asks them to registered agents, including the applet
if available; the applet now supports an external-ui-mode in which it
spawns a VPN-specific authentication dialog binary that doesn't have a
graphical component and contains the knowledge about which secrets are
missing, how to retrieve them from keyrings and how to ask them to
users; the binary then returns through stdout a keyfile that describes
the known secrets, their values returned from keyrings and the labels
to present to users. The applet uses this information to show a GTK
dialog (only when necessary) and returns the secrets to NM.

Currently when using the external-ui-mode and the auth-dialog binary
returns secrets with the ShouldAsk key set to false, the applet
doesn't show a GTK dialog and returns an empty response. This breaks
when the auth-dialog returns secrets from keyrings as they are
dropped.

This commit changes the handling of secrets in external-ui-mode: now
we first build a list of known secrets and, if necessary, populate the
dialog with the ones that must be asked. After the dialog quits we
update the secret values in the list from the dialog and return all
secrets, including ones not asked.

#59
https://gitlab.freedesktop.org/NetworkManager/NetworkManager/issues/193

Fixes: bce8d0a7 ("applet/vpn-request: add external UI mode")

#71

Fix the following warning:

 "Failed to set text 'ab&cd' from markup due to error parsing markup:
 Error on line 1: Entity did not end with a semicolon; most likely you
 used an ampersand character without intending to start an entity —
 escape ampersand as &"

Don't first enable markup (with the old label) and then set the text;
do both with one call to gtk_label_set_markup().

https://gitlab.freedesktop.org/NetworkManager/NetworkManager/issues/244

Without gtk_widget_show_all() the popup menu for password flags is
displayed empty when using GTK3.

Fixes: 2ded57d9 ("utils: support GTK4")

It was added in meson 0.50 while we require 0.46. Furthermore it
defaults to true when a install directory is specified and so it can
be omitted.

#70

This is EXPERIMENTAL EXPERIMENTAL EXPERIMENTAL. GTK4 is not released yet
and things will change in incompatible ways.

A new library called "libnma-gtk4" is added. The non-libnma parts
(applet, editor) are not ported.

There are few reasons for merging this early and none of them is
shipping the library to the users at this point:

* To allow bringup of GTK4-enabled VPN plugins
* Do not block GTK4 enablement of GNOME Control Center
* Help avoid using features that will be difficult to port from GTK3

All the test programs work, they emit a fair amount of warnings about
properties that were removed in GTK4. To deal with those, we'd need to
build different GResources for GTK3 and GTK4. The differencies seem
small enough to the point the conversion could be automated. TODO,
doesn't block testing.

Tested to build with gtk+-4.0 = 3.94.0, but even the pkg-config name
changed since in GTK4 git master. Did I say this is EXPERIMENTAL and
there will be incompatible changes?

Cover a trivial difference with an ifdef.

Just paint over the differencies with a few ifdefs.

It's going away in Gtk4. We don't do show_all() anyway -- just hide the
widgets by default.

Essentially to get the thing to compile for now.

Replaces Gtk4's gtk_window_activate_default. This probably deserves a
better solution, but it will have to do for now.

There's no no-show-all in GTK4. Using it in GTK3 is perhaps also a bad
idea, but we can't change that now without potentially breaking things
at this point.