The last processor editing commit broke various parts of the
cpu editor window.
Most noticeably switching between the default and non-default
cpu types dealt with empty values wrongly and tried to
delete the 'cputype' property rather than the cputype
portion of the cpu property string.
The backup and iothread options are now boolean types
internally instead of strings and need to be treated as
such.
Added a parseBoolean() function to deal with this with an
optional default value to use for undefined values.
The default for an undefined backup value is true.
API changes in ExtJS6 involved here:
* 'checkchange' event passes now the record index instead of the whole record
* record.fields is now an array of Field objects, before it was an Object
The use of this field raise an error with ExtJS6 because it is not
unique inside the application.
Removing this is safe, because we never query the Comment column
with an id.
This fixes a nasty bug where an initial click on the ViewSelector
trigger would display the picker but hide it immediatly afterwards.
What was happening behind the scene, is that without queryMode 'local',
the store bound to the picker was loaded on each picker display.
Loading the store would then trigger our onLoad() override in ToolKit.js
which called setValue(), which fired a bunch of unwanted events
including 'collapse'.
It is safe here not to call onLoad() to set an initial value, because
the initial value is set in initComponent() (value: groupdef[0][0])
This makes the code more homogeneous and allows us to use the ExtJS
class autoloader in the future if we want (autoloader idea: ExtJS tries to load the
JS class file over HTTP based on the class name, no need to maintain a list
of includes)
Some fields need to be processed by initConfig(),
which will be run before initComponent() in the
component lifecycle.
This fix the problem of the Selector default value
whihc was not loading.
If set, Diffie-Hellman parameters in PEM format are loaded
from the given path. Otherwise, the built-in 'skip2048'
group is used.
Also fix some typos in the man page.
Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
AnyEvent uses a built-in DH group defined as 'schmorp1539'
by default, which seems to trigger the bug in [1] for every
attempt of accessing the web GUI using IE11 and TLS1.2. By
switching to a bigger default DH group ('skip2048'), the
bug seems to be gone (or trigger sufficiently rarely).
1: http://engineering.imvu.com/2015/01/27/the-case-of-the-page-cant-be-displayed-intermittent-selenium-test/
Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
If we pass true to the removeAll() , ExtJS6 will try to erase the leafes of a tree,
and then try to to iterate throuch these leaves in onNodeRemove() ( this obviously do not work)
The true parameter does not seem to be needed anyway anymore.
After a call to removeAll, the childNodes[] property of the rootNode is empty
Instead of extending the framework ComboBox, we extend the abstract
parent class Picker, so we don't rely on private methods ( and hopefull survive the
next ExtJS upgrade )
Usability improvement:
The selection model is now a checkboxmodel, which makes more clear to the user that
we can select more than one element in the list.
Options such as:
osd mount options xfs = -i size=2048 -n size=16k
were parsed as:
key: 'osd mount options xfs = -i size=2048 -n size'
value: '16k'
This is due to the greedy .* in the beginning. Making it
non-greedy fixes this.
Fixes#851
The Picker of the ComboBox, has by default a maxHeight of 300 px. If the list of elements is bigger that 300 pixels, then a scrollbar will be added.
This behaviour was broken with ExtJS4, but now works correctly with ExtJS6.
This fixes the problem that this ComboGrid was displayed with an empty value
( similar the to the Keyboard Selection and Realm Selection in the login window)
