GitLab

Apparently <numeric> header is included by something indirectly for us on
Linux (and FreeBSD and Windows CI's). Anyway doesn't hurt to add the
include.

Reported by Alexander Lohnau:
!646 (comment 382424)

GIT_SILENT

When KFile::Directory mode changes, we need to update the dir,
otherwise initially we would be showing dirs and files even when
dirOnlyMode() is true.

With the use of embedded json metadata, the exec value is the same as the file the metadata was loaded from

In case of conflict in i18n, keep the version of the branch "ours"
To resolve a particular conflict, "git checkout --ours path/to/file.desktop"

Laurent Montel authored Jan 18, 2022 and Christoph Cullmann committed Jan 18, 2022

suggestion

GIT_SILENT

Based on kfilecustomdialogtest_gui. This should help with an upcoming MR:
!713

GIT_SILENT

The code here basically wants to be signaled when the network configuration
changes so that it can re-read WPAD (Web Proxy Auto-Detection) settings. In
Qt5 QNetworkConfiguration::Defined was used, which isn't available any more
in Qt6.

Instead use QNetworkInformation::Reachability, and check for
QNetworkInformation::Reachability::Local/Site.

Note this hasn't been tested as I don't have any proxy setups.

GIT_SILENT

NO_CHANGELOG

This way it gets the same QT_PLUGIN_PATH variable exported like the other
unittests.

Hopefully exporting that var isn't going to be necessary any more after[1],
but using ecm_add_test() is more in-line with the other unittests.

[1] kcoreaddons!164

Q6DateTime::fromString() will return an invalid datetime if the "GMT" part
is included, so chop it off, Qt uses UTC by default anyway, so the datetime
we get back isn't affected by this chopping.

NO_CHANGELOG

In Qt6 QVector::size() returns a qsizetype, on my system it's a qint64,
in UDSEntry load()/save() we beging de/serializing the data with "size"
which is a quint32, so use that explicitly here to make the test pass.

The usage here is a simple "isOnline()".

GIT_SILENT

- Fix some include directives
- Fix QtConcurrent::run usage, the signature has changed in Qt6
- Q6ByteArray::size() returens qsizetype, aka long long, adjust sprintf()
  syntax
- int can't be implicitly converted to QChar any more, create a QChar
  explicitly
- Q6StringBuilder is confused when concatenating QByteArray's, help it
- Q6FileInfo constructor that takes a QString is now explicit, use
  QDirIterator::fileInfo() directly
- QString::size() returns a qsizetype, cast it to int when creating a
  QJSValue
- Use QSortFilterProxyModel::filterRegularExpression(), QRegularExpression
  vs. QRegExp (deprecated in Qt6)
- Create a QKeySequence explicitly, implicit conversion from int is gone

GIT_SILENT

GIT_SILENT

viewOptions() is gone.

GIT_SILENT

GIT_SILENT

This was built with:
-DQT_MAJOR_VERSION=6 \
-DEXCLUDE_DEPRECATED_BEFORE_AND_AT=5.90.0 \
-DKF_DISABLE_DEPRECATED_BEFORE_AND_AT=0x055a00

This makes the code a bit simpler to read, shorter and more resilient against messed up json types

Unlike the other kded modules, the affected plugins by this commit are installed in the "kf5/kiod" namespace.
Those are considered KIO-internal and do not get displayed in the "Background Services" KCM.
Consequently we do not need the name or description. Keeping them around would only increase build
times and binary sizes. Also the translators might put effort into sth. that is not visible to any users.

The ServiceTypes are unused and consequently removed.

Qt5 treats those as regular structs and uses the alias as their name. Qt6
distinguishes those two and thus D-Bus argument marshaling fails when
encountering unnamed structs.

Using regular named structs fixes this.

See also https://phabricator.kde.org/T15158.

This test tried to move /etc/paswd into the trash which was an
incredibly dangerous thing to try to do, because if that move
operation ever succeeded, it would brick your whole system!

With polkit support being merged, this test now does just that every
time it's run.

This commit is an emergency removal so that people don't brick their
systems by running KIO's test suite anymore.

Task: https://phabricator.kde.org/T15109

Implements the spec as in https://specifications.freedesktop.org/desktop-entry-spec/desktop-entry-spec-latest.html#dbus
DBus activation will only be used if the application supports it and
no options such as DeleteTemporaryFiles or suggested file names are
specified that cannot be implemented using the spec.

This is misleading as they don't react to clicks either.
Unfortunately, the `KFilePlacesEventWatcher` had to be extended so that
we could trigger a repaint for the relevant cells as needed.

This obviously meant to be `MiddleButton` :)

Make place items a little less tall but instead give section headers
more padding.

Move it into a shared function called from context menu and inline.

Volker Krause authored Jan 05, 2022 and David Faure committed Jan 10, 2022

See https://mail.kde.org/pipermail/kde-frameworks-devel/2022-January/120608.html
for the background on this.

This is considered compatible as both prefixes are included in the build
metadata, and local tests with Plasma and PIM confirm that so far.

I picked KIO as a test subject as kio_version.h is one of the most widely
used version headers. I'd suggest we try this here for 5.91 and if that
doesn't cause any fallout roll this out to all frameworks.