The same issues, GNOME Shell 3.24.0, Ubuntu GNOME 17.04 amd64:

Gjs-Message: JS WARNING: [resource:///org/gnome/shell/ui/main.js 317]: reference to undefined property global.stage
Gjs-Message: JS WARNING: [resource:///org/gnome/shell/ui/layout.js 217]: reference to undefined property global.window_group
Gjs-Message: JS WARNING: [resource:///org/gnome/shell/ui/layout.js 899]: reference to undefined property this.monitors[this.findIndexForActor(...)]
Gjs-Message: JS WARNING: [resource:///org/gnome/shell/ui/osdMonitorLabeler.js 59]: reference to undefined property Meta.MonitorManager
Gjs-Message: JS WARNING: [resource:///org/gnome/shell/ui/panel.js 1107]: reference to undefined property indicator.menu._openChangedId
Gjs-Message: JS WARNING: [resource:///org/gnome/gjs/modules/tweener/tweener.js 514]: reference to undefined property obj.delay
current session already has an ibus-daemon.
Gjs-Message: JS WARNING: [resource:///org/gnome/shell/ui/slider.js 34]: reference to undefined property St.GenericAccessible
Gjs-Message: JS WARNING: [resource:///org/gnome/shell/ui/search.js 432]: reference to undefined property provider.isRemoteProvider
Gjs-Message: JS WARNING: [resource:///org/gnome/shell/misc/ibusManager.js 196]: reference to undefined property this._engines[id]
Gjs-Message: JS WARNING: [resource:///org/gnome/shell/ui/status/network.js 1885]: reference to undefined property this._mainConnectionStateChangedId
Gjs-Message: JS WARNING: [resource:///org/gnome/shell/ui/panel.js 342]: reference to undefined property this.menu.actionGroup
Gjs-Message: JS WARNING: [resource:///org/gnome/shell/ui/dash.js 760]: reference to undefined property oldApps[oldIndex]
Gjs-Message: JS WARNING: [resource:///org/gnome/shell/ui/appDisplay.js 1298]: reference to undefined property this.name
Gjs-Message: JS WARNING: [resource:///org/gnome/shell/ui/workspace.js 1229]: reference to undefined property this._reservedSlotWindow

*** Bug 783896 has been marked as a duplicate of this bug. ***

Created attachment 353949 [details] [review]
Initialize properties in _init()

While we've always considered it good style to initialize JS properties,
some code that relies on uninitialized properties having an implicit
value of 'undefined' has slipped in over time. The updated SpiderMonkey
version used by gjs now warns when accessing those properties, so we
should make sure that they are properly initialized to avoid log spam,
even though all warnings addressed here occur in conditionals that
produce the correct result with 'undefined'.

Created attachment 353950 [details] [review]
Don't use 'undefined' as boolean

Make sure we pass a proper false value instead of undefined to
avoid warnings.

Created attachment 353951 [details] [review]
dash: Fix some JS warnings

We currently use "array[index]" to test whether an array has an
element at index before using it. However nowadays gjs warns about
accessing non-existent array elements, so the test itself already
produces a warning. Avoid this by checking the array length before
using an index to access an element.

Created attachment 353952 [details] [review]
remoteMenu: Add actionGroup accessor

To avoid recreating the app menu unnecessarily, the panel checks
whether the menu's current actionGroup already matches the target
one. However as the menu's actionGroup property is currently private,
the test always fails, whoops.

Created attachment 353953 [details] [review]
panel: Fix a JS warning

We need to track the open-status of indicator menus, but don't want
to hook up signals more than once, so we check for the handler ID
we store on the object. As the property is only defined once we did
set up the signal connection, this check now logs a warning. We
can avoid it by checking for the existence of the property rather
than a particular value.

Created attachment 353954 [details] [review]
history: Fix bogus return value

We only use lastItem() to reset the history index to the end, so
nobody noticed the utter nonsense in the return value until gjs
started to warn about it. As we don't actually use the value
anywhere, we could just remove it, but the function name implies
that an item is returned, so fix it to behave as advertised.

Created attachment 353955 [details] [review]
ibusManager: Fix a warning

Don't try to access a non-existent engine - it probably makes sense to
use Map() instead of a plain object to track engines in the future, but
for now just add an additional check to shut up a warning.

Created attachment 353956 [details] [review]
layout: Range-check index before array lookup

findMonitorForActor() may be called before the layoutManager gets
to initialize monitors, so make sure the monitor index is in range
to avoid a warning.

Created attachment 353957 [details] [review]
windowManager: Fix a warning

The destroy effect currently relies on a monkey-patched property
added from the map effect. However on X11 it is possible that we
did no map animation for a window that is destroyed when the shell
was restarted or had taken over from another WM. Just use the real
MetaWindow property to avoid a warning in that case.

Created attachment 353958 [details] [review]
calendar: Fix access to undefined property

Don't call a parent method before chaining-up _init() to
avoid a warning about access to an undefined property.

Created attachment 353959 [details] [review]
notificationDaemon: Avoid access to undefined property

Avoid a warning when assigning from an undefined property by making
sure to pass a proper null value instead.

Created attachment 353960 [details] [review]
dateMenu: Set _delegate on non-menu-item menu child

PopupMenu needs special-treatment of certain types of menu items,
which it determines via children's _delegate property. However as
the calendar drop-down is very unmenu-ish, we use regular actors
rather than PopupMenuItems and the missing _delegate property
triggers a warning. Just add it as the bare minimum to make
PopupMenu happy.

Created attachment 353961 [details] [review]
shellDBus: Avoid access to undefined properties

Don't rely on an implicit undefined value of unset parameters to
shut up some warnings.

The attached patches should fix most of the JS warnings that are triggered by gnome-shell as seen in my logs (and mostly matching the ones on this bug).

(In reply to Georges Basile Stavracas Neto from comment #0)
> Gjs-Message: JS WARNING: [resource:///org/gnome/shell/ui/main.js 317]:
> reference to undefined property global.stage

The ShellGlobal singleton is initialized before handing over control to javascript, so why all its GObject properties trigger a warning is a complete mystery to me. Any hints appreciated.


> [resource:///org/gnome/shell/ui/osdMonitorLabeler.js 59]: reference to
> undefined property Meta.MonitorManager

Same for valid classes in properly imported typelib namespaces ...

Created attachment 353966 [details] [review]
Initialize properties in _init()

Found another case ...

Review of attachment 353950 [details] [review]:

ok

Review of attachment 353951 [details] [review]:

seems fine and more readable

Review of attachment 353952 [details] [review]:

right

Review of attachment 353953 [details] [review]:

++

Review of attachment 353954 [details] [review]:

obviously

Review of attachment 353955 [details] [review]:

indeed

Review of attachment 353956 [details] [review]:

this looks fine but callers will still blow up since only one deals with a null monitor

Review of attachment 353957 [details] [review]:

might as well just remove the JS property and always use the real one, no?

Review of attachment 353958 [details] [review]:

::: js/ui/calendar.js
@@ +765,2 @@
         this.setUseBodyMarkup(notification.bannerBodyMarkup);
+        this.setBody(notification.bannerBodyText);

nit: it would also work to keep the body text instead of null above and avoid this second call

Review of attachment 353959 [details] [review]:

ok

Review of attachment 353960 [details] [review]:

sure

Review of attachment 353961 [details] [review]:

++

Review of attachment 353966 [details] [review]:

all lgtm

Created attachment 355513 [details] [review]
layout: Skip strut computation in the no-monitor case

It's possible for updateRegions() to be called before monitors have
been properly initialized. Instead of throwing an error in that case,
just skip the strut computation (that doesn't make sense anyway without
a monitor).


(In reply to Rui Matos from comment #24)
> this looks fine but callers will still blow up since only one deals with a
> null monitor

Yeah, it's not meant to fix the no-monitor case, just to guard against any calls during initialization. The only caller that looks like we might run into on startup (although I haven't seen it in the logs) is the one in _updateRegions(), so probably worth handling it.

(In reply to Rui Matos from comment #25)
> Review of attachment 353957 [details] [review] [review]:
> 
> might as well just remove the JS property and always use the real one, no?

It's used in the 'notify::window-type' handler to update the dimming of the parent window if either the old or new window type was "attached-modal-dialog". I don't see how we can do the "old window type" part without tracking it ourselves ...


(In reply to Rui Matos from comment #26)
> Review of attachment 353958 [details] [review] [review]:
> 
> ::: js/ui/calendar.js
> @@ +765,2 @@
>          this.setUseBodyMarkup(notification.bannerBodyMarkup);
> +        this.setBody(notification.bannerBodyText);
> 
> nit: it would also work to keep the body text instead of null above and
> avoid this second call

Oh, true - nice catch!

Review of attachment 355513 [details] [review]:

looks fine

Review of attachment 353957 [details] [review]:

(In reply to Florian Müllner from comment #32)
> It's used in the 'notify::window-type' handler to update the dimming of the
> parent window if either the old or new window type was
> "attached-modal-dialog". I don't see how we can do the "old window type"
> part without tracking it ourselves ...

Aha, indeed, I misread that code

Review of attachment 353956 [details] [review]:

++

Attachment 353950 [details] pushed as e2f88d6 - Don't use 'undefined' as boolean
Attachment 353951 [details] pushed as 2a01606 - dash: Fix some JS warnings
Attachment 353952 [details] pushed as 755698a - remoteMenu: Add actionGroup accessor
Attachment 353953 [details] pushed as 28d0983 - panel: Fix a JS warning
Attachment 353954 [details] pushed as b50d714 - history: Fix bogus return value
Attachment 353955 [details] pushed as f33560f - ibusManager: Fix a warning
Attachment 353956 [details] pushed as 1ba014d - layout: Range-check index before array lookup
Attachment 353957 [details] pushed as 32fdc22 - windowManager: Fix a warning
Attachment 353958 [details] pushed as b73c96d - calendar: Fix access to undefined property
Attachment 353959 [details] pushed as d815c5d - notificationDaemon: Avoid access to undefined property
Attachment 353960 [details] pushed as c1758c4 - dateMenu: Set _delegate on non-menu-item menu child
Attachment 353961 [details] pushed as ccaae5d - shellDBus: Avoid access to undefined properties
Attachment 353966 [details] pushed as 358133e - Initialize properties in _init()
Attachment 355513 [details] pushed as 4e57b45 - layout: Skip strut computation in the no-monitor case

I still see some warnings that fall in the two categories mentioned in comment #16; I'll leave this open for now in the hope that we'll figure those out as well ...

Created attachment 355667 [details] [review]
Remove excess parameter to g_app_info_launch()

Commit 39a840e2c3d added an additional parameter to shell_app_launch().
When adjusting callers, the parameter was also added accidentally to
calls of the confusingly similar g_app_info_launch() ...
Remove those to fix some warnings.

Review of attachment 355667 [details] [review]:

yep

Comment on attachment 355667 [details] [review]
Remove excess parameter to g_app_info_launch()

Attachment 355667 [details] pushed as 6b71b79 - Remove excess parameter to g_app_info_launch()

With gnome-shell 3.26.1, only a few warnings are left:

Okt 25 22:25:07 hostname gnome-shell[899]: JS WARNING: [resource:///org/gnome/shell/ui/main.js 315]: reference to undefined property "MetaStage"
Okt 25 22:25:07 hostname gnome-shell[899]: JS WARNING: [resource:///org/gnome/shell/ui/layout.js 221]: reference to undefined property "MetaWindowGroup"
Okt 25 22:25:07 hostname gnome-shell[899]: JS WARNING: [resource:///org/gnome/shell/ui/osdMonitorLabeler.js 59]: reference to undefined property "MetaDBusDisplayConfigSkeleton"
Okt 25 22:25:07 hostname gnome-shell[899]: JS WARNING: [resource:///org/gnome/shell/ui/slider.js 38]: reference to undefined property "CallyActor"
Okt 25 22:25:08 hostname gnome-shell[899]: JS WARNING: [resource:///org/gnome/gjs/modules/tweener/tweener.js 540]: reference to undefined property "isSpecialProperty"
Okt 25 22:25:09 hostname gnome-shell[899]: JS WARNING: [resource:///org/gnome/shell/ui/windowManager.js 1471]: reference to undefined property "MetaWindowX11"
Okt 25 22:37:19 hostname gnome-shell[1597]: JS WARNING: [resource:///org/gnome/shell/ui/popupMenu.js 715]: reference to undefined property "_delegate"

Repeated warnings have been removed from this list.

(In reply to Christian Stadelmann from comment #40)
> With gnome-shell 3.26.1, only a few warnings are left:
> 
> Okt 25 22:25:07 hostname gnome-shell[899]: JS WARNING:
> [resource:///org/gnome/shell/ui/main.js 315]: reference to undefined
> property "MetaStage"

Those are the kinds of warnings mentioned in comment #16 - still no idea what's wrong there.


> Okt 25 22:25:08 hostname gnome-shell[899]: JS WARNING:
> [resource:///org/gnome/gjs/modules/tweener/tweener.js 540]: reference to
> undefined property "isSpecialProperty"

That's not gnome-shell - I've submitted https://gitlab.gnome.org/GNOME/gjs/merge_requests/5


> Okt 25 22:37:19 hostname gnome-shell[1597]: JS WARNING:
> [resource:///org/gnome/shell/ui/popupMenu.js 715]: reference to undefined
> property "_delegate"

That's likely a non-PopupMenuItem-actor added to a menu without setting the _delegate property like we do here:
https://git.gnome.org//browse/gnome-shell/tree/js/ui/dateMenu.js#n490

As far as I can see we don't stuff custom actors into a menu anywhere else, so I suspect that this is triggered by an extension.

(In reply to Florian Müllner from comment #41)
> (In reply to Christian Stadelmann from comment #40)
> > With gnome-shell 3.26.1, only a few warnings are left:
> > 
> > Okt 25 22:25:07 hostname gnome-shell[899]: JS WARNING:
> > [resource:///org/gnome/shell/ui/main.js 315]: reference to undefined
> > property "MetaStage"
> 
> Those are the kinds of warnings mentioned in comment #16 - still no idea
> what's wrong there.

I had another look today, and finally figure it out \o/

Created attachment 371621 [details] [review]
clutter: Don't reference invalid pc in gir

Cally is built into clutter itself rather than exposed as a separate
library.

Created attachment 371622 [details] [review]
window-group: Remove undefined function declaration

Ever since the function has been made public, its name has been
meta_actor_is_untransformed() ...

Created attachment 371623 [details] [review]
build: Introspect some more types

While MetaStage, MetaWindowGroup and MetaDBusDisplayConfigSkeleton don't
appear explicitly in the public API, their gtypes are still exposed via
meta_get_stage_for_screen(), meta_get_*window_group_for_screen() and
MetaMonitorManager's parent type. Newer versions of gjs will warn about
undefined properties if it encounters a gtype without introspection
information, so expose those types to shut up the warnings.

Created attachment 371624 [details] [review]
build: Include Cally in St introspection

CallyActor is exposed indirectly via StAccessible's parent type,
so add the dependency to shut up a gjs warning.

Also filed as https://gitlab.gnome.org/GNOME/mutter/merge_requests/100 and https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/94.

Created attachment 371784 [details] [review]
build: Introspect some more types

While MetaStage, MetaWindowGroup and MetaDBusDisplayConfigSkeleton don't
appear explicitly in the public API, their gtypes are still exposed via
meta_get_stage_for_screen(), meta_get_*window_group_for_screen() and
MetaMonitorManager's parent type. Newer versions of gjs will warn about
undefined properties if it encounters a gtype without introspection
information, so expose those types to shut up the warnings.

Attachment 371621 [details] pushed as 920dc9e - clutter: Don't reference invalid pc in gir
Attachment 371622 [details] pushed as b380aa7 - window-group: Remove undefined function declaration
Attachment 371784 [details] pushed as 28eff93 - build: Introspect some more types

Attachment 371624 [details] pushed as a49fb90 - build: Include Cally in St introspection

*** Bug 730087 has been marked as a duplicate of this bug. ***