Alexander Mikhaylenko authored Dec 02, 2021 and Marge Bot committed Dec 02, 2021

GTK4 removes it and leaves margin-top etc. Use them directly.

Part-of: <!1038>

Alexander Mikhaylenko authored Dec 02, 2021 and Marge Bot committed Dec 02, 2021

Use margin-start instead.

Part-of: <!1038>

Alexander Mikhaylenko authored Dec 02, 2021 and Marge Bot committed Dec 02, 2021

It's gone in GTK4, use margins.

Part-of: <!1038>

Alexander Mikhaylenko authored Dec 02, 2021 and Marge Bot committed Dec 02, 2021

They will go away in GTK4. The default on GtkFrame is fine, on scrolled
window add .frame style class instead.

Delete random label_xalign as well, they are unused.

Part-of: <!1038>

Alexander Mikhaylenko authored Dec 02, 2021 and Marge Bot committed Dec 02, 2021

This will go away in GTK4, use .flat style class instead.

Part-of: <!1038>

Alexander Mikhaylenko authored Dec 02, 2021 and Marge Bot committed Dec 02, 2021

A lot of places where we used these properties wasn't necessary. In some
cases it worked but would mess up labels in GTK4.

Part-of: <!1038>

Alexander Mikhaylenko authored Dec 02, 2021 and Marge Bot committed Dec 02, 2021

Part-of: <GNOME/epiphany!1038>

Alexander Mikhaylenko authored Dec 02, 2021 and Marge Bot committed Dec 02, 2021

There was a period where GtkGrid looked like it's going to be the future,
along with a misleading mention that boxes are going to be removed. In
practice, that was about HBox and VBox, not Box. Meanwhile, this layout has
no reason whatsoever to use grid and it just complicates it.

Part-of: <!1038>

Alexander Mikhaylenko authored Dec 02, 2021 and Marge Bot committed Dec 02, 2021

It's gone in GTK4, use separate hexpand/vexpand as needed.

Part-of: <!1038>

Alexander Mikhaylenko authored Dec 01, 2021 and Marge Bot committed Dec 02, 2021

webkit_file_chooser_request_select_files() is transfer none, we still need
to free the array.

Part-of: <!1038>

Alexander Mikhaylenko authored Dec 01, 2021 and Marge Bot committed Dec 02, 2021

It was from the time when this was using a template. It doesn't anymore,
and it's one less line to adapt for GTK4.

Part-of: <GNOME/epiphany!1038>

Alexander Mikhaylenko authored Dec 01, 2021 and Marge Bot committed Dec 02, 2021

Stop relying on GtkContainer for this, it will go away in GTK4.

Part-of: <GNOME/epiphany!1038>

Alexander Mikhaylenko authored Dec 01, 2021 and Marge Bot committed Dec 02, 2021

It can do things such as show a spinner, a secondary label or actions.

Remove all of that, only leave what we actually use.

Part-of: <!1038>

Alexander Mikhaylenko authored Dec 01, 2021 and Marge Bot committed Dec 02, 2021

In GTK4 gtk_box_pack_start() is gone, replaced with gtk_box_append(). More
importantly, child properties are gone and pack_start() allows to set them.

With this in mind, stop using anything other than their default values:
FALSE, TRUE, 0. Everything else can be done with halign, valign, hexpand,
vexpand and margins.

gtk_box_pack_end() is gone as well. Replace with pack_start() where
possible (which involves rearranging the order in which widgets are added),
and the remaining uses at this point can be replaced with gtk_box_prepend().

Part-of: <GNOME/epiphany!1038>

Alexander Mikhaylenko authored Dec 01, 2021 and Marge Bot committed Dec 02, 2021

Part-of: <!1038>

Sophie Herold authored Sep 24, 2021 and Marge Bot committed Dec 01, 2021

Part-of: <!1016>

Quentin PAGÈS authored Dec 01, 2021 and Administrator committed Dec 01, 2021

(cherry picked from commit 3f824fce)

Phaedrus Leeds authored Nov 30, 2021 and Marge Bot committed Dec 01, 2021

Part-of: <!1037>

Carlos Garcia Campos authored Nov 23, 2021 and Michael Catanzaro committed Nov 30, 2021

It seems CI doesn't support it

Part-of: <!1028>

This should improve scrolling performance and also performance with 4k
monitors. Well, it should improve performance in general. It may of
course expose bugs to be fixed. Now seems like the time to tackle them.

Part-of: <!1028>

Users can input whatever they want for the name of the web app, as long
as it is valid UTF-8. Since a72d21d7 we
have passed this along straight to GApplication without sanitizing it to
ensure it can safely be used in a GApplication ID. If the user decides
to use any character in the app name besides Latin alphanumeric
characters, we are doomed. This has never worked properly, but
previously all we saw were criticals from GApplication as it dropped our
bogus app ID. But since WebKitGTK 2.32, WebKit actually refuses to start
a sandboxed subprocess without a valid app ID, as it should.

The complication here is that we cannot simply decide to create better
app IDs for existing web apps: the app ID must exactly match the desktop
file, and we want it to match the profile directory name too. So let's
assume that non-Latin web apps were previously broken, hope that users
delete any such broken web apps rather than leak them on disk, and
handwave away migration issues by altering the app ID only if it cannot
safely be used in a GApplication. Existing non-broken web apps continue
to use the same app IDs as before. Newly-created web apps with only
Latin characters follow the old format for IDs, while newly-created web
apps with non-approved characters will follow the new format. Old broken
web apps remain broken forever, until the user decides to delete them.

The old app ID format is:
`org.gnome.Epiphany.WebApp-<normalized-name>-<checksum>`

The new format is:
`org.gnome.Epiphany.WebApp-<checksum>`

The ideal format would be:
`org.gnome.Epiphany.WebApp_<checksum>`

because hyphens in app IDs are deprecated, but let's not take on that
battle today.

Note that in this commit message, "app ID" refers to the GApplication
ID, which is actually confusingly different from the EphyWebApplication
"app ID."

This commit also replaces the concept of "program name" with the
GApplication ID. It was very confusing to have "name" and "program name"
be two completely different things. Now we can have "ID" and
"GApplication ID" be two different things instead. Yay.

Fixes #1627

See also: #1626

Part-of: <!1032>

Part-of: <!1035>

Part-of: <!1035>

This code has been very fragile in the past, so the similar rework when
closing a tab will be in a separate commit.

Copypaste the original function - without gtk_dialog_run() sharing it
between the 2 code paths isn't worth it.

Part-of: <!1035>

Part-of: <!1035>

Part-of: <!1035>

Part-of: <!1035>

It also needed a manual modal=true, turns out gtk_dialog_run() does that
implicitly.

Part-of: <!1035>

This one needed more refactoring than others.

Part-of: <!1035>

Part-of: <!1035>

Part-of: <!1035>

Part-of: <!1035>

gtk_dialog_run() and gtk_native_dialog_run() will go away in GTK4. Rework
everything to use the response signal.

Part-of: <!1035>

It's already set by the fact it's the default response.

Part-of: <!1035>

This is a sad one - the only reason revealer exists is to mask the fact
GtkActionBar hide animation is broken. It's fixed in gtk4 and that revealer
will go away, but just need a stepping stone for now.

Part-of: <GNOME/epiphany!1035>

Part-of: <GNOME/epiphany!1035>

Part-of: <!1035>

They are gone in GTK4.

Part-of: <!1035>

This looks like legacy, not sure what it was even used for.

Part-of: <!1035>

They are gone in GTK4.

Part-of: <!1035>