Resolves#1313
What this commit does:
* Whenever the Apply button is pressed, and if the save was successful, then the Apply button is disabled.
* Each subwidget used by EditEntryWidget has now a signal called `widgetUpdated` that is emitted when the widgets' internal content changes. The EditEntryWidget subscribes to that signal to know when to enable the Apply button (by calling `entryUpdated()`).
* There are some views that are not isolated in their own widgets (`m_advancedUi`, for example) so in those cases I invoked `entryUpdated()` directly whenever I detected an update:
* some updates occur directly in a Qt widget like when editing the text of a QLineItem, so in that case I connected the widget's signals directly to the `entryUpdated()` slot.
* some updates occur in EditEntryWidget, so in those cases the invocation to `entryUpdated()` is made as soon as the change is detected (for example when the user has confirmed an action in a dialog).
A known problem: there are some situations when the Apply button will get enabled even if there are no changes, this is because the app changes the value of a field by itself so it's considered an update (for example, clicking on the "Reveal" button changes the text shown in a text field).
The solution to this can be a bit complicated: disabling temporarily the `entryUpdated()` whenever the app is going to do an action with such side-effects.
So I preferred to let the Apply button get enabled in those cases.
Ignoring closeEvent signals that the application doesn't want the widget
to be closed. This may cause unwanted shutdown interruption. There's no
difference between close() and hide() unless WA_DeleteOnClose is set, so
it's better to just accept the event if we're hiding the window
afterwards anyway.
Resolves#856
Fix/work around KeePassXC flickering and not restoring from tray on some Linux systems, which happens if the window is hidden and minimized by code at the same time (see issue #1595).
* Tests: Speed up AutoType testing
Decrease default autotype delay to 1 to improve test suite speed by
seconds. This shaves multiple seconds off the whole test suite. In some
cases, the largest part.
Also, initialize config just creating the test instance, just in case
that it ever depends on the configuration values at that point already.
* Tests: Speed up Kdbx4 testing
This speeds up the Kdbx4 tests by using parameters optimized for speed
for the key derivation functions. On an i7-6700K the tests run close to
50% faster with this change (about 1.5s vs. 3s).
- Fix unnecessary automatic upgrade to KDBX 4.0 and prevent challenge-response key being stripped [#1568]
- Abort saving and show an error message when challenge-response fails [#1659]
- Support inner stream protection on all string attributes [#1646]
- Fix favicon downloads not finishing on some websites [#1657]
- Fix freeze due to invalid STDIN data [#1628]
- Correct issue with encrypted RSA SSH keys [#1587]
- Fix crash on macOS due to QTBUG-54832 [#1607]
- Show error message if ssh-agent communication fails [#1614]
- Fix --pw-stdin and filename parameters being ignored [#1608]
- Fix Auto-Type syntax check not allowing spaces and special characters [#1626]
- Fix reference placeholders in combination with Auto-Type [#1649]
- Fix qtbase translations not being loaded [#1611]
- Fix startup crash on Windows due to missing SVG libraries [#1662]
- Correct database tab order regression [#1610]
- Fix GCC 8 compilation error [#1612]
- Fix copying of advanced attributes on KDE [#1640]
- Fix member initialization of CategoryListWidgetDelegate [#1613]
- Fix inconsistent toolbar icon sizes and provide higher-quality icons [#1616]
- Improve preview panel geometry [#1609]
The 'text' property of the QAction gets mangled by KDE when it adds its
own accelerator shortcuts. But the data property is ours. Use that for
keying instead.
Signed-off-by: Steven Noonan <steven@uplinklabs.net>
Similar to KeePass2 most elements are checked for a
Protected="True"
attribute. Previously, files KDBX 3.1 files with protected Binary
entries, as seen in the wild, could not be read (mangled passwords and
file attachments).
When a field value is of the form ...{REF:...}...
Then:
* Auto-type does not work with that field
* References to this field are not properly resolved
This commit fixes both issues.
Building with -flto caught the fact that we were ignoring the return
value of readMagicNumbers(), which potentially left the value of 'sig2'
uninitialized.
Signed-off-by: Steven Noonan <steven@uplinklabs.net>
