Age | Commit message (Collapse) | Author |
|
Follow the advice in #7049
|
|
When double-clicking a network resource in the devtools, one
get AttributeError: 'WebEnginePage' object has no attribute 'view'.
|
|
|
|
|
|
FindFlags -> FindFlag
int() argument must be a string, a bytes-like object or a number, not 'FindFlag'
I didn't look for other occurances
|
|
The developer tools only saves preferences to disk, and reads them from
disk, if the profile of the page they are drawing to isn't
off-the-record. So if you want it to remember your dark mode preference
the page hosting the inspector needs to be linked to an on-the-record
profile. So we need to create a new page linked to a specific profile
and then set that on the view.
Options at this point:
1. (in `__init__()` always make a new page linked to the qute default
profile and add that to the view
2. delay creating the view and page until `inspect()` is called and get
a pointer to the profile of the page being inspected
(1) is actually what we have done on previous Qt versions because there
the inspector was started with WebEngine's default, global, profile,
which used to be on-the-record. I'm not sure if there are any specific
ramifications to having the inspector data persistent when the page it
is attached to isn't, but there might be stuff like per-domain
preferences saved in that case which would be an information leak.
So I've went with (2) even though it is probably going to be more
disruptive to the tests. In the future if we re-use inspectors across
different tabs/windows again we can make a new page on demand if the
profile of the inspected page differs form the initial one. For now I've
put an assert there because I am lazy.
It's a bit awkward having most of the setup in `inspect()` and not
`__init__()`. `_init_inspector()` and `inspect()` are only ever called
once right beside each other though (now) so we could just pass the page
to `__init__()` and get rid of `inspect()`.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Qt6 has switched to the global default profile being off-the-record,
which is not the default for qutebrowser.
Restore the previous default object name (although we always point it
elsewhere anyway so that name should appear in any file paths).
ref https://bugreports.qt.io/browse/QTBUG-66068
|
|
ba6055d5d38a3f3 included some hasty fixups that caused a couple of
places to the from ApplicationWorld to 1, instead of 2. This change
mostly reverts that can calls .value in a couple of places since the
WebEngine APIs wants ints for some reason.
|
|
|
|
|
|
`QVariant.Type` has moved to `QMetaType.Type`[1][] and QMeta.Type
doesn't work with int().
`QImage(':/icons/qutebrowser-64x64.png')` yields and empty QImage. This
is not fixed but things don't crash because of it anymore. For instance
the "Title & Body" test on https://web-push-book.gauntface.com/demos/notification-examples/
`QtWebEngineCore.QWebEnginePage.Feature` doesn't work with int(), so add
it to the maps twice.
`for s in scripts` was from a previous hack
[1]: https://www.qt.io/blog/whats-new-in-qmetatype-qvariant
|
|
FindFlag doesn't work with int() either, use .value.
FindTextResult doesn't act like a bool, call `numberOfMatches()`.
Undo some disabled stuff that has been fixed upstream
Emit a bool, not the FindTextResult object.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
See #5904
|
|
|
|
|
|
|
|
https://blog.chromium.org/2020/02/protecting-users-from-insecure.html
https://therecord.media/firefox-follows-chrome-and-prepares-to-block-insecure-downloads/
|
|
|
|
|
|
|
|
|
|
Looks like the Edge one doesn't work anymore.
Reverts db13e5283e2d20e83d348f15ea0f46c92e9f3e29 and reintroduces the Firefox UA
removed in e010afd3a20a86639396a9c844abfea7b23cc67a (but updated).
Closes #5182.
|
|
|
|
|
|
|
|
|
|
comment
|
|
|
|
This reverts commit e2c5fe6262564d9d85806bfa9d4486a411cf5045.
See https://mail.python.org/archives/list/python-dev@python.org/thread/ZMC67QA2JVQJSWSFWRS6IM6ZX4EK277G/#LSTMFAPSPD3BGZ4D6HQFODXZVB3PLYKF
|
|
Database class
|
|
|
|
|
|
|
|
|