Re: patch for gtk-vnc not picked up by bots

["Thomas H.P. Andersen" <phomes gmail com>]

On Wed, Dec 15, 2010 at 14:38, Thomas H.P. Andersen <phomes gmail com> wrote:
> On Wed, Dec 15, 2010 at 14:01, Piñeiro <apinheiro igalia com> wrote:
>> From: "Thomas H.P. Andersen" <phomes gmail com>
>>
>>> It seems the patch[1] for gtk-vnc is not being picked up by the build
>>> bots. The builds are still failing with the specific lines removed in
>>> the patch. On a local "jhbuild update gtk-vnc" the patch is applied as
>>> expected.
>>
>> Are you sure that this is the problem? If you take a look to the
>> update log on each slave [1][2][3], you can see that each time the
>> buildbot compiles a module it makes a updateone on this module:
>>
>>  argv: ['jhbuild', '--moduleset=gnome-apps-3.0', 'bot', '--step', 'updateone', u'gtk-vnc'
>>
>> And, AFAIU, this is the same solution that you are suggesting.
>
> The problem is that the build step fails with the same lines removed
> by the patch:
> vncdisplay.c: In function 'motion_event':
> vncdisplay.c:600: error: 'GdkDrawable' undeclared (first use in this function)
> vncdisplay.c:600: error: (Each undeclared identifier is reported only once
> vncdisplay.c:600: error: for each function it appears in.)
> vncdisplay.c:600: error: 'drawable' undeclared (first use in this function)
> vncdisplay.c:600: warning: implicit declaration of function 'GDK_DRAWABLE'
> vncdisplay.c:600: warning: nested extern declaration of 'GDK_DRAWABLE'
> vncdisplay.c: In function 'on_cursor_changed':
> vncdisplay.c:1211: warning: passing argument 1 of
> 'gdk_window_get_display' makes pointer from integer without a cast
> /home/jhbuild/sid-ia32/build/include/gtk-3.0/gdk/gdkwindow.h:533:
> note: expected 'struct GdkWindow *' but argument is of type 'int'
>
> The patch:
> diff --git a/src/vncdisplay.c b/src/vncdisplay.c
> index 33d2623..fdb56b5 100644
> --- a/src/vncdisplay.c
> +++ b/src/vncdisplay.c
> @@ -629,9 +629,9 @@ static gboolean motion_event(GtkWidget *widget,
> GdkEventMotion *motion)
>
>        /* Next adjust the real client pointer */
>        if (!priv->absolute) {
> -               GdkDrawable *drawable = GDK_DRAWABLE(gtk_widget_get_window(widget));
> -               GdkDisplay *display = gdk_drawable_get_display(drawable);
> -               GdkScreen *screen = gdk_drawable_get_screen(drawable);
> +               GdkWindow *window = gtk_widget_get_window(widget);
> +               GdkDisplay *display = gdk_window_get_display(window);
> +               GdkScreen *screen = gdk_window_get_screen(window);
>                int x = (int)motion->x_root;
>                int y = (int)motion->y_root;
>
> @@ -1260,7 +1260,7 @@ static void on_cursor_changed(VncConnection
> *conn G_GNUC_UNUSED,
>        }
>
>        if (cursor) {
> -               GdkDisplay *display =
> gdk_drawable_get_display(GDK_DRAWABLE(gtk_widget_get_window(GTK_WIDGET(obj))));
> +               GdkDisplay *display = gtk_widget_get_display(GTK_WIDGET(obj));
>                GdkPixbuf *pixbuf = gdk_pixbuf_new_from_data(vnc_cursor_get_data(cursor),
>                                                             GDK_COLORSPACE_RGB,
>                                                             TRUE, 8,
>
> so I assumed that the patch was not applied correctly. I was just
> guessing that the reason could be that the patch has wrong line
> offsets. On my local it successfully applies the patch but complains
> about the offset.
>
>>> (I noticed that the patch has a wrong offset for the lines but it
>>> succeeds to apply anyway. Just wondering it this could be a reason.)
>>>
>>> [1]http://git.gnome.org/browse/jhbuild/tree/patches/gtk-vnc.drawable.patch
>>
>> BR
>>
>> [1] http://build.gnome.org/builders/gtk-vnc-bxlug-sid/builds/1066/steps/gtk-vnc%20update/logs/stdio
>> [2] http://build.gnome.org/builders/gtk-vnc-fedora-64/builds/6/steps/gtk-vnc%20update/logs/stdio
>> [3] http://build.gnome.org/builders/gtk-vnc-cosimoc-deb-i686/builds/2/steps/gtk-vnc%20update/logs/stdio
>>
>> ===
>> API (apinheiro igalia com)
>>
>

I have figured out what is going on. This is a bug (I guess?) in
jhbuild. Patches are only applied for tarballs when they are
downloaded/unpacked. The problem here is that the patch was added in
jhbuild after the bots had already downloaded the tarball. (Except for
cosimoc, but it is failing for a missing dep)

Workaround could be to nuke the checkout of gtk-vnc and it should
start working. I will file a bug for jhbuild later tonight.