All commands will be offered as completions for the <command> argument
of :bind.
Due to the way completers parse the command line, the following
bind --mode caret j
will throw off completions as 'caret' is treated as a positional arg in
terms of the argument count for completions.
In the above example, completion will be triggered for 'j' and no
completions will be given for the actual command.
bind --mode=caret j will complete correctly, though completions are not
filtered by the given mode.
I attempted an approach to filter the commands based on the mode but it
ended up being messy and flaky.
The <command> arg is now optional. If omitted, :bind prints the current
binding as a message. If --mode is given, the binding for that mode is
printed.
Allow a variable amount of whitespace for rgb, rgba, hsv, and hsva
strings in the config.
Previously only 'rgb(0, 0, 0)' was allowed. Now things like
'rgb(0,0,0)' are permitted.
The repeated 3-digit segments of the regexes were separated out to
reduce repetition and line length.
If the process emitted error() and then finished(), we already set
self._filepath to None and did other cleanup.
Instead we do the file reading inside _cleanup and call that from
on_process_error and on_process_finished.
Issue #1412
When passing --cachedir="" on the command line, standarddir.cache()
returns None, which stands for "deactivate cache" and has to be
properly handled in DiskCache.__init__() (i.e. don't pass it to
os.path.join)
This worked fine with Python 3.5 but causes a circular import which is
hard to break with Python 3.4.
The original solution was to do @pyqtSlot(object), but that doesn't work
with PyQt 5.6 anymore...
When using :tab-prev/:tab-next (or :tab-focus which uses :tab-next
internally) immediately after the last tab, those functions could be
called with 0 tabs open, which caused a ZeroDivisionError when trying to
do % 0.
Fixes#1448.
Allows to use ctrl+d to close tabs from the buffer completion widget
when they are selected. Respects current tab settings like whether you
can close the last tab in a window.
Had to change the `rebuild()` method to use `setData()` when possible
because the selection was being lost if the whole model was being rebuilt.
Current problems are:
1) When opening a new window while you already the tab completion open on
one window a category is added for the new window but new rows in
that category aren't picked up. Interesting if you open a third
window then close the second window the completion display is now
correct... I can see that the model is being updated correctly but I
am not sure why that isn't propagating to the view. Not sure whether
it is worth looking into (further) either.
2) Bit of duplication of code, it iterates over the window registry
twice. Could put everything in one loop but then that would be
dependant on the current behaviour of the `tab_closed` signal being
called with the relevant `tabbed_browser` still existing but with the
`shutting_down` flag set.
3) I'm still using just the one `rebuild()` method and removing items from
the end then calling `setData` on everything rather than having special
`on_tab/window_closed` methods (or partial functions) that delete the
actual corresponding item. Because if I did that I would also have to
special case tab moves etc.
Avoid a crash when undoing twice on the default page with last-close set to
default-page.
This was caused by logic to reuse the current tab if it is on the default page
and has no history. The fix is using openurl rather than removeTab/tabopen.
Mark two callbacks with pyqtSlot as appropriate.
Return False instead of raising NotImplementedException to avoid pylint
identifying MarkKeyParser as abstract.
Ensure MarkKeyParser implements on_keyconfig_changed, so it doesn't fail when
rebinding a key. It doesn't have keybindings, so the implementation is just
`pass`.
This also fixes a few flake8 style errors.
- Fix a docstring copy-paste
- Add own name/copyright date to new file
- Simplify a bdd expression (no need for regex)
- Scroll to a pixel position in a single operation
Rather than binding each set of local marks to a tab, bind them to a
url. Strip the fragment from the url, as two pages that differ only in
fragment are likely the same page.
Automatically set the special "'" mark when jumping.
jump-mark "'" will jump to the last position before the previous jump.
A jump could be navigating via a link, jumping to another mark, or
scrolling by percentage (e.g. gg or G).
set-mark <key> saves your current scroll position as mark <key>.
jump-mark <key> jumps to the position previously set for mark <key>.
If <key> is lowercase, it is local to the current tab. Each tab has its
own set of lowercase marks.
If <key> is uppercase, it is global across tabs, and stores a url and a
scroll position. Jumping to an uppercase mark navigates to that url,
then scrolls to the saved position.
Resolves#310.
* upstream/master: (327 commits)
Remove unused import
tox: Update Werkzeug to 0.11.8
Regenerate authors
Use __file__ instead of sys.argv[0]
Regenerate authors
Make update_3rdparty.py install correctly when run from any directory
Open command line urls explicitly.
tox: Update Werkzeug to 0.11.6
Move qutebrowser.rcc to misc/
Regenerate resources
Fix CHANGELOG/link in README
New qutebrowser logo!
www: Add releases link
Release v0.6.1
release checklist: Clarify how to build on Windows
Make sure the cheatsheet PNG is included in sdist
Fix cheatsheet link URL in quickstart
Mark segfault on exit in test_smoke as xfail
Add a xfail test for #797
Add missing file
...
Conflicts:
tests/integration/features/hints.feature
Under some circumstances I can't reproduce (switching/turning off
monitors?) it seems it's possible that SessionManager.save gets called
with last_window=True, without on_last_window_closed being called.
This might be to one of the Qt screen management bugs fixed in Qt 5.6,
which would explain why I can't reproduce it.
Instead of crashing, let's log the error and not save the session.
This makes it possible to jump to the very last tab, as opposed to the
last focused tab, by using -1 as the index. Generally negative indexes
are counted from the end.
Solves issue #1166
This was needed before there was editor.ExternalEditor as there were
various commands which needed to access the editor object.
Since this is encapsulated in ExternalEditor now, no need to keep a
reference to the object around.
The edit-url command opens a url (by default, the current url) in the
user's external editor and navigates to the result when the editor is
closed. This makes it easy to tweak the current url to navigate within
a site.
`edit-url` accepts the same flags as `open` (e.g. -t will open in a new
tab.
One may provide a url as an argument to create a shortcut to
pre-populate part of a url and allow filling in the rest.
There is no default keybinding.
Resolves#1261.
Searching for that error doesn't turn up many helpful results, but it
seems to be harmless and shown when downloading a file - it's also new
in Qt 5.6 it seems, so let's just ignore it.
The previous fix didn't work in situations where the web view was
actually focused, but had no focused element (like about:blank).
The new fix always works, and even is a lot simpler!
Fixes#504.
When a download is finished with `removed-finished-download` set to a
delay, it's removed via a singleshot QTimer.
However, when the window was closed in the meantime, the slot still was
executed by Qt, even though the DownloadManager was already deleted.
Fixes#1242
Those were added in #443, inspired by luakit.
However, all other bindings follow dwb's defaults, and dwb uses `gt` for
showing buffers. To be consistent, let's rebind gt to show :buffer.
`buffer` takes either a tab index or a string and focuses the specified
tab. The index can be of the form [0-9]+ which will switch to the
relevant tab in the current window or [0-9]+/[0-9]+ (that is
win_id/index) which will focus the specified window before switching
tabs. If a string is passed the list of open tabs across all windows is
sorted based on title and url (just like in the completion widget) and
the top result is selected.
Issue #1334
The problem was that there were too few slashes. On Linux, absolute
paths start with /, so
file:// + /home
gives file:///home, which is a valid path. On windows however, absolute
paths start with a drive letter, so
file:// + C:/Users
gives file://C:/Users, which is parsed as "host C, path Users", which is
why it could be written as file://c/Users (strip out the empty "port"),
giving us an invalid path.
The solution is to add the third slash in the template, and strip the
redundant slash on unix systems.
Additionally, this fixes a bug where navigating from '/home/' to the
parent directory would give '/home' instead of '/'
Fix for #676
It removes the target of the link, as to prevent the website to overrule
the user. I guess the following things should be done:
- add setting to enable/disable this behaviour
- and/or add "hint all current"
Only the first one would be easiest. The second one requires us to keep track
of the original target. I should open a pull request for discussion.
We can get UndefinedError when a new function got added to the jinja
env (and gets called from a template) and the user did update the
on-disk templates but not restart qutebrowser yet.
In this case, let's show a special error page to the user and tell them
to do :report in the unlikely case it's actually a bug.
Fixes#1362.
See #1360.
Before, in fonts -> _monospace we preferred Liberation Mono to Courier and
Courier New. Unfortunately, that font looks horrible on Windows if it's
installed, so let's avoid using it if possible.
On Linux, hopefully Courier New/Courier isn't available, and Liberation Mono
will get picked up instead (or one of the other alternatives).
* upstream/master: (22 commits)
Regenerate authors.
Only run geolocation tests on CI
Switch to flake8-docstrings with pydocstyle
Fix lint
Move pylint plugins to an installed package.
Include pytest-xvfb properly in frozen tests
tests: Handle trailing / in wait_for_load_finished.
Fix lint.
tox: Upgrade hypothesis to 3.0.2.
tests: Add first end-to-end test for hints.
hints: Log the used hint chars
bdd: use quteproc.wait_for_load_finished.
Remove xvfbwrapper from freeze_tests.py.
Regenerate authors.
Combine launch/crash time into one section.
Split long lines.
Switch to pytest-xvfb. Fixes#1309.
no ellipsis is inserted in big windows
Regenerate authors.
fixes#1308
...
Issues #1269, #866
qutebrowser would crash when XDG_DOWNLOAD_DIR was set to some
non-absolute value (which should not happen, but it can) and
"storage -> download-dir" was empty, since when the user didn't give an
absolute filename, even the joined path of download_dir() (i.e.
XDG_DOWNLOAD_DIR in this case) and the filename was not absolute either.
Since the path was not absolute, create_full_filename returned None,
which meant that os.path.basename(self._filename) raised an exception.
Now we display an error message and fall back to $HOME.
Issue #1214
Now uses a sensible filename for data: links instead of the whole base64
content. For PDF.js, it even uses the correct pdf filename.
TODO: Produces "QPainter:🔚 Painter ended with 2 saved states" while
running the tests here (Arch Linux):
CPython: 3.5.1
Qt: 5.5.1, runtime: 5.5.1
PyQt: 5.5.1
Otherwise the stacktrace might be confusing since it will show the
FileNotFoundException as the causing error, which is not true (it just
happens to be the last checked place).
The .path attribute was added so that we still have the requested path
in the error log.
See #1280.
For some reason, when comparing the repr in the two processes, we get different
results on OS X and Windows:
- expected: "fünf"
- "f\xfcnf" coming back from the subprocess on OS X
- "fnf" on Windows
Instead we're comparing the json dump now, which should be more predictable.
There are a lot of problems and flakiness with using a real clipboard.
Instead we now have a :debug-set-fake-clipboard command to set a text, and use
logging when getting the contents.
Fixes#1285.
Before we raised QtValueError (via qtutils.ensure_valid), but maybe there are
more callers out there which call fuzzy_url with an empty input - and it makes
more sense to raise InvalidUrlError which gets displayed to the user than
raising QtValueError which is more like an assertion.
I tried to set my editor to `termite -e "vim -f {}"`, termite being a
pretty cool and light terminal I use within my i3wm Arch linux box.
So when I open my editor I want it to launch a terminal with Vim inside
instead of GVim for various reasons.
The validation rejected this at first because it was looking for '{}'
inside ['foo', 'bar', 'baz {}'], essentially. So I need it to look
inside the sub-strings, not just the list.
Then after validation I need to perform the '{}' replacement inside the
sub-string too, not just replacing the whole string.
WebHistory now has a clear() method which is also a command
(history-clear) which clears the qutebrowser history using the new
lineparser clear() method and emits a cleared signal.
The completion model urlmodel connects to the WebHistory.cleared signal
and clears its history category completion list.
I am adding this as a temporary fix before #58 or #1051 get implemented.
It seems that unlike Gecko, WebKit does not support undo/redo operations
when the textarea's `value` attribute is changed directly. Fortunately
there is a WebKit-specific workaround using textInput event.
References:
* http://stackoverflow.com/a/7554295
* http://help.dottoro.com/ljuecqgv.php
The Shift+Ins key should arguably insert primary selection, not the
clipboard selection as every Qt program does. This commit makes it
possible via the hidden paste-primary command (enabled by default).
Unfortunately QtWebKit does not provide any straightforward way to
insert text at cursor position into editable fields, so we work around
this by executing a JavaScript snippet - inspired by this SO answer:
http://stackoverflow.com/a/11077016
This logged an error on Windows:
ERROR misc networkmanager:on_authentication_required:269 Unable to read the netrc file
Traceback (most recent call last):
File "c:\python34\Lib\netrc.py", line 27, in __init__
file = os.path.join(os.environ['HOME'], ".netrc")
File "C:\Users\florian\buildbot\slave\win8\build\.tox\py34\lib\os.py", line 633, in __getitem__
raise KeyError(key) from None
KeyError: 'HOME'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "C:\Users\florian\buildbot\slave\win8\build\qutebrowser\browser\network\networkmanager.py", line 262, in on_authentication_required
net = netrc.netrc()
File "c:\python34\Lib\netrc.py", line 29, in __init__
raise OSError("Could not find .netrc: $HOME is not set")
Since this case is pretty common, we don't want to log it - and checking the
variable beforehand is easier than parsing the exception message.
This should fix the failing tests on Windows.
Otherwise, on OS X we got the same SSL error logged twice as on_ssl_errors is
called twice. This means the tests only marked one as expected, and it failed
because of the other one.
Since 2a705e2eb6 non-specialized config types are
String. However, String had an overloaded complete() which defaulted to
returning None.
Now we use the normal complete() which relies on valid_values if completions
isn't given instead.
Fixes#1223.
- The paste command will now open one tab/window per url if multiple
URLs (separated by newline) are present in the clipboard
- Adds the tests for the new multitab functionality
- Changes test/integration/conftest.py to be able to insert newlines in
the clipboard for the test
