Display info icon for Initialized/Inactive extension state

[TanmayPatil105]

Updates

Related issues

#646

[oscfdezdz]

Thank you for looking into this! I think we don't need to add some of the states as tooltips since we show icons with tooltips with more detailed explanation. These would exclude ERROR and OUT_OF_DATE states, a screenshot of the latter:

Also, it seems that the states of the extensions in the different distributions don't work the same. In Fedora, for the same extensions you have shown, I see:

I'm not quite sure why this happens, but it makes things a little more difficult for us to try to improve how the status is displayed to the user.

[oscfdezdz]

Reviewing all states, the transient states ACTIVATING and DEACTIVATING are also dispensable, the switch transition is self-explanatory and short enough that it isn't necessary. And with ACTIVE state, the active switch with the accent color highlighting also makes it redundant.

That leaves us with INACTIVE and INITIALIZED, while the former should be clear because of the deactivated switch, it seems that it isn't the case on all systems. Still, with these two states being the only ones causing confusion, I'd prefer that we add the tooltip to an icon as in the case of ERROR and OUT_OF_DATE, if that's ok with you.

You can see the details of the implementation of the icons in that same source file and in exm-extension-row.blp. Using an icon like info-outline-symbolic should do it. Although it still remains to find out, as mentioned before, why the states are different depending on the system. I'll try to look into it in the next few days.

[TanmayPatil105]

Also, it seems that the states of the extensions in the different distributions don't work the same. In Fedora, for the same extensions you have shown, I see:

Yeah, that's weird. I'll look into that too!!

Thank you for looking into this! I think we don't need to add some of the states as tooltips since we show icons with tooltips with more detailed explanation. These would exclude ERROR and OUT_OF_DATE states, a screenshot of the latter:

Reviewing all states, the transient states ACTIVATING and DEACTIVATING are also dispensable, the switch transition is self-explanatory and short enough that it isn't necessary. And with ACTIVE state, the active switch with the accent color highlighting also makes it redundant.

That makes sense.

You can see the details of the implementation of the icons in that same source file and in exm-extension-row.blp. Using an icon like info-outline-symbolic should do it. Although it still remains to find out, as mentioned before, why the states are different depending on the system. I'll try to look into it in the next few days.

Sounds good. I'll update the patch.
Thanks!

[TanmayPatil105]

Not related to this PR but while viewing an Extension details the reviews are getting leaked onto the terminal.

tanmay@tanmay-lenovo:~/Desktop/projects/gnome/extension-manager$ ./_build/src/extension-manager
<p>Works on Nobora 39</p>
Root Node discovered: html
Ignored element html
Ignored element body
<p>Not working on Ubuntu 22.04.4</p>
Root Node discovered: html
Ignored element html
Ignored element body
<p>Works!!!!!<br>Thank You!!!!!!!!!!!!!!!!!!!!!!</p>
Root Node discovered: html
Ignored element html
Ignored element body
<p>Woks great!</p>
Root Node discovered: html
Ignored element html
Ignored element body
<p>Great Extension, now working on Gnome 46, thank you</p>
Root Node discovered: html
Ignored element html
Ignored element body

That must be due to a stray printf?

[oscfdezdz]

Yes, that's #361. I looked at it a while ago and it's from text-engine.

[TanmayPatil105]

[TanmayPatil105]

@oscfdezdz Please take a look!

[oscfdezdz]

It's looking great, thank you very much! 😄

While we find the cause of the difference of states between systems, you can check if the INACTIVE state and GtkSwitch properties active is true and state is false (systems like yours) and the INITIALIZED state and GtkSwitch properties active is true and state is false (systems like mine). That will display only in extensions like "Control Blur Effect On Lockscreen" or "Customize Clock on Lock Screen" on both systems without difference.

[TanmayPatil105]

INACTIVE:

[TanmayPatil105]

I figured something out:
After booting up, the state of the extensions is INITIALIZED.
If we lock and then unlock the screen, it becomes INACTIVE.

Could you check if this is reproducible?

[oscfdezdz]

That's it, I can reproduce it. There are no differences then.

Sorry if I didn't explain myself well in my last comment 😅 I meant that we should check the status and active properties of the switch in the code to show the icon, in those specific cases. Now that I think about it again, I don't think it would be necessary, with the enabled variable it should be enough.

In the screenshot you sent there were manually disabled extensions, such as DING or Ubuntu Dock that have the icon visible and should not.

Now that you have figured out why the states were different, something like:

gtk_widget_set_visible (GTK_WIDGET (self->info_icon),
                            (state == EXM_EXTENSION_STATE_INITIALIZED
                            || state == EXM_EXTENSION_STATE_INACTIVE)
                            && enabled);

should work, I haven't test it.

I'd also now use one tooltip instead (could be set in blp file directly) with an understandable message for all users. I was thinking of:

This extension could be activated in session modes such as the login screen or the lock screen

[TanmayPatil105]

Done, Thanks!

[oscfdezdz]

Thank you for raising the issue and fixing it! From my side all is good, I'll let @mjakeman have the last word.

[mjakeman]

Hi @TanmayPatil105, thanks for the contribution!

I think as a general philosophy, the app shouldn't have a state where a tooltip/icon is needed to explain why a "core feature" does not work intuitively. We should either hide the toggle, or bind it to a different property. In this case, the toggle really should be bound to enabled rather than active going off the issue you raised for shell.

However I also think that perfect is the enemy of good, and this will definitely be useful in the short term.

I'm happy to merge for now, with a long term view to redo how we handle enabled/active :)