# Copyright 2014-2021 Florian Bruhin (The Compiler) # # This file is part of qutebrowser. # # qutebrowser is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # qutebrowser is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # You should have received a copy of the GNU General Public License # along with qutebrowser. If not, see . """Download manager.""" import io import os.path import shutil import functools import dataclasses from typing import Dict, IO, Optional from qutebrowser.qt.core import pyqtSlot, pyqtSignal, QTimer, QUrl from qutebrowser.qt.widgets import QApplication from qutebrowser.qt.network import QNetworkRequest, QNetworkReply, QNetworkAccessManager from qutebrowser.config import config, websettings from qutebrowser.utils import message, usertypes, log, urlutils, utils, debug, objreg, qtlog from qutebrowser.misc import quitter from qutebrowser.browser import downloads from qutebrowser.browser.webkit import http from qutebrowser.browser.webkit.network import networkmanager @dataclasses.dataclass class _RetryInfo: request: QNetworkRequest manager: QNetworkAccessManager class DownloadItem(downloads.AbstractDownloadItem): """A single download currently running. There are multiple ways the data can flow from the QNetworkReply to the disk. If the filename/file object is known immediately when starting the download, QNetworkReply's readyRead writes to the target file directly. If not, readyRead is ignored and with self._read_timer we periodically read into the self._buffer BytesIO slowly, so some broken servers don't close our connection. As soon as we know the file object, we copy self._buffer over and the next readyRead will write to the real file object. Attributes: _retry_info: A _RetryInfo instance. _buffer: A BytesIO object to buffer incoming data until we know the target file. _read_timer: A Timer which reads the QNetworkReply into self._buffer periodically. _reply: The QNetworkReply associated with this download. _autoclose: Whether to close the associated file when the download is done. Signals: adopt_download: Emitted when a download is retried and should be adopted by the QNAM if needed. arg 0: The new DownloadItem """ adopt_download = pyqtSignal(object) # DownloadItem def __init__(self, reply, manager): """Constructor. Args: reply: The QNetworkReply to download. """ super().__init__(manager=manager, parent=manager) self.fileobj: Optional[IO[bytes]] = None self.raw_headers: Dict[bytes, bytes] = {} self._autoclose = True self._retry_info = None self._reply = None self._buffer = io.BytesIO() self._read_timer = usertypes.Timer(self, name='download-read-timer') self._read_timer.setInterval(500) self._read_timer.timeout.connect(self._on_read_timer_timeout) self._url = reply.url() self._init_reply(reply) def _create_fileobj(self): """Create a file object using the internal filename.""" assert self._filename is not None try: # pylint: disable=consider-using-with fileobj = open(self._filename, 'wb') except OSError as e: self._die(e.strerror) else: self._set_fileobj(fileobj) def _do_die(self): """Abort the download and emit an error.""" self._read_timer.stop() if self._reply is None: log.downloads.debug("Reply gone while dying") return self._reply.downloadProgress.disconnect() self._reply.finished.disconnect() self._reply.errorOccurred.disconnect() self._reply.readyRead.disconnect() with qtlog.hide_qt_warning('QNetworkReplyImplPrivate::error: Internal ' 'problem, this method must only be called ' 'once.'): # See https://codereview.qt-project.org/#/c/107863/ self._reply.abort() self._reply.deleteLater() self._reply = None if self.fileobj is not None: pos = self.fileobj.tell() log.downloads.debug(f"File position at error: {pos}") try: self.fileobj.close() except OSError: log.downloads.exception("Error while closing file object") if pos == 0: # Emtpy remaining file filename = self._get_open_filename() log.downloads.debug(f"Removing empty file at {filename}") try: os.remove(filename) except OSError: log.downloads.exception("Error while removing empty file") def _init_reply(self, reply): """Set a new reply and connect its signals. Args: reply: The QNetworkReply to handle. """ self.done = False self.successful = False self._reply = reply reply.setReadBufferSize(16 * 1024 * 1024) # 16 MB reply.downloadProgress.connect(self.stats.on_download_progress) reply.finished.connect(self._on_reply_finished) reply.errorOccurred.connect(self._on_reply_error) reply.readyRead.connect(self._on_ready_read) reply.metaDataChanged.connect(self._on_meta_data_changed) reply.redirected.connect(self._on_redirected) self._retry_info = _RetryInfo(request=reply.request(), manager=reply.manager()) if not self.fileobj: self._read_timer.start() # We could have got signals before we connected slots to them. # Here no signals are connected to the DownloadItem yet, so we use a # singleShot QTimer to emit them after they are connected. if reply.error() != QNetworkReply.NetworkError.NoError: QTimer.singleShot(0, lambda: self._die(reply.errorString())) @pyqtSlot(QUrl) def _on_redirected(self, url): if self._reply is None: log.downloads.warning(f"redirected: REPLY GONE -> {url}") else: log.downloads.debug(f"redirected: {self._reply.url()} -> {url}") def _do_cancel(self): self._read_timer.stop() if self._reply is not None: self._reply.finished.disconnect(self._on_reply_finished) self._reply.abort() self._reply.deleteLater() self._reply = None if self.fileobj is not None: self.fileobj.close() self.cancelled.emit() @pyqtSlot() def retry(self): """Retry a failed download.""" assert self.done assert not self.successful assert self._retry_info is not None # Not calling self.cancel() here because the download is done (albeit # unsuccessfully) self.remove() self.delete() new_reply = self._retry_info.manager.get(self._retry_info.request) new_download = self._manager.fetch(new_reply, suggested_filename=self.basename) self.adopt_download.emit(new_download) def _get_open_filename(self): filename = self._filename if filename is None: filename = getattr(self.fileobj, 'name', None) return filename def url(self) -> QUrl: # Note: self._reply is deleted when the download finishes return self._url def origin(self) -> QUrl: if self._reply is None: return QUrl() origin = self._reply.request().originatingObject() try: return origin.url() except AttributeError: # Raised either if origin is None or some object that doesn't # have its own url. return QUrl() def _ensure_can_set_filename(self, filename): if self.fileobj is not None: # pragma: no cover raise ValueError("fileobj was already set! filename: {}, " "existing: {}, fileobj {}".format( filename, self._filename, self.fileobj)) def _after_set_filename(self): self._create_fileobj() def _ask_confirm_question(self, title, msg, *, custom_yes_action=None): yes_action = custom_yes_action or self._after_set_filename no_action = functools.partial(self.cancel, remove_data=False) url = 'file://{}'.format(self._filename) message.confirm_async(title=title, text=msg, yes_action=yes_action, no_action=no_action, cancel_action=no_action, abort_on=[self.cancelled, self.error], url=url) def _ask_create_parent_question(self, title, msg, force_overwrite, remember_directory): assert self._filename is not None no_action = functools.partial(self.cancel, remove_data=False) url = 'file://{}'.format(os.path.dirname(self._filename)) message.confirm_async(title=title, text=msg, yes_action=(lambda: self._after_create_parent_question( force_overwrite, remember_directory)), no_action=no_action, cancel_action=no_action, abort_on=[self.cancelled, self.error], url=url) def _set_fileobj(self, fileobj, *, autoclose=True): """Set the file object to write the download to. Args: fileobj: A file-like object. """ assert self._reply is not None if self.fileobj is not None: # pragma: no cover raise ValueError("fileobj was already set! Old: {}, new: " "{}".format(self.fileobj, fileobj)) self.fileobj = fileobj self._autoclose = autoclose try: self._read_timer.stop() log.downloads.debug("buffer: {} bytes".format(self._buffer.tell())) self._buffer.seek(0) shutil.copyfileobj(self._buffer, fileobj) self._buffer.close() if self._reply.isFinished(): # Downloading to the buffer in RAM has already finished so we # write out the data and clean up now. self._on_reply_finished() else: # Since the buffer already might be full, on_ready_read might # not be called at all anymore, so we force it here to flush # the buffer and continue receiving new data. self._on_ready_read() except OSError as e: self._die(e.strerror) def _set_tempfile(self, fileobj): self._set_fileobj(fileobj) def _finish_download(self): """Write buffered data to disk and finish the QNetworkReply.""" assert self._reply is not None assert self.fileobj is not None log.downloads.debug("Finishing download...") if self._reply.isOpen(): self.fileobj.write(self._reply.readAll()) if self._autoclose: self.fileobj.close() self.successful = self._reply.error() == QNetworkReply.NetworkError.NoError self._reply.close() self._reply.deleteLater() self._reply = None self.finished.emit() self.done = True log.downloads.debug("Download {} finished".format(self.basename)) self.data_changed.emit() @pyqtSlot() def _on_reply_finished(self): """Clean up when the download was finished. Note when this gets called, only the QNetworkReply has finished. This doesn't mean the download (i.e. writing data to the disk) is finished as well. Therefore, we can't close() the QNetworkReply in here yet. """ if self._reply is None: return self._read_timer.stop() self.stats.finish() log.downloads.debug("Reply finished, fileobj {}".format(self.fileobj)) if self.fileobj is not None: # We can do a "delayed" write immediately to empty the buffer and # clean up. self._finish_download() @pyqtSlot() def _on_ready_read(self): """Read available data and save file when ready to read.""" if self.fileobj is None or self._reply is None: # No filename has been set yet (so we don't empty the buffer) or we # got a readyRead after the reply was finished (which happens on # qute://log for example). return if not self._reply.isOpen(): raise OSError("Reply is closed!") try: self.fileobj.write(self._reply.readAll()) except OSError as e: self._die(e.strerror) @pyqtSlot('QNetworkReply::NetworkError') def _on_reply_error(self, code): """Handle QNetworkReply errors.""" if code == QNetworkReply.NetworkError.OperationCanceledError: return if self._reply is None: error = "Unknown error: {}".format( debug.qenum_key(QNetworkReply, code)) else: error = self._reply.errorString() self._die(error) @pyqtSlot() def _on_read_timer_timeout(self): """Read some bytes from the QNetworkReply periodically.""" assert self._reply is not None if not self._reply.isOpen(): raise OSError("Reply is closed!") data = self._reply.read(1024) if data is not None: self._buffer.write(data) @pyqtSlot() def _on_meta_data_changed(self): """Update the download's metadata.""" if self._reply is None: return self.raw_headers = {} for key, value in self._reply.rawHeaderPairs(): self.raw_headers[bytes(key)] = bytes(value) def _uses_nam(self, nam): """Check if this download uses the given QNetworkAccessManager.""" assert self._retry_info is not None running_nam = self._reply is not None and self._reply.manager() is nam # user could request retry after tab is closed. retry_nam = (self.done and (not self.successful) and self._retry_info.manager is nam) return running_nam or retry_nam class DownloadManager(downloads.AbstractDownloadManager): """Manager for currently running downloads. Attributes: _networkmanager: A NetworkManager for generic downloads. Class attributes: _MAX_REDIRECTS: The maximum redirection count. """ # Same as many browsers # https://fetch.spec.whatwg.org/#http-redirect-fetch # https://source.chromium.org/chromium/chromium/src/+/main:net/url_request/url_request.h;l=97;drc=3c19a2edb96d3d5b56a7481349a357fdbdf8ecf0 # https://stackoverflow.com/questions/9384474/in-chrome-how-many-redirects-are-too-many _MAX_REDIRECTS = 20 def __init__(self, parent=None): super().__init__(parent) self._networkmanager = networkmanager.NetworkManager( win_id=None, tab_id=None, private=config.val.content.private_browsing, parent=self) @pyqtSlot('QUrl') def get(self, url, cache=True, **kwargs): """Start a download with a link URL. Args: url: The URL to get, as QUrl cache: If set to False, don't cache the response. **kwargs: passed to get_request(). Return: The created DownloadItem. """ if not url.isValid(): urlutils.invalid_url_error(url, "start download") return None req = QNetworkRequest(url) user_agent = websettings.user_agent(url) req.setHeader(QNetworkRequest.KnownHeaders.UserAgentHeader, user_agent) if not cache: req.setAttribute(QNetworkRequest.Attribute.CacheSaveControlAttribute, False) # Needed for Qt 5, default on Qt 6 # We don't set this on the QNAM because QtWebKit handles redirects manually. req.setAttribute( QNetworkRequest.Attribute.RedirectPolicyAttribute, QNetworkRequest.RedirectPolicy.NoLessSafeRedirectPolicy, ) req.setMaximumRedirectsAllowed(self._MAX_REDIRECTS) return self.get_request(req, **kwargs) def get_mhtml(self, tab, target): """Download the given tab as mhtml to the given DownloadTarget.""" assert tab.backend == usertypes.Backend.QtWebKit from qutebrowser.browser.webkit import mhtml if target is not None: mhtml.start_download_checked(target, tab=tab) return suggested_fn = utils.sanitize_filename(tab.title() + ".mhtml") filename = downloads.immediate_download_path() if filename is not None: target = downloads.FileDownloadTarget(filename) mhtml.start_download_checked(target, tab=tab) else: question = downloads.get_filename_question( suggested_filename=suggested_fn, url=tab.url(), parent=tab) question.answered.connect(functools.partial( mhtml.start_download_checked, tab=tab)) message.global_bridge.ask(question, blocking=False) def _get_suggested_filename(self, request): """Get the suggested filename for the given request.""" filename_url = request.url() if request.url().scheme().lower() == 'data': # We might be downloading a binary blob embedded on a page or even # generated dynamically via javascript. If we happen to know where it's # coming from, we can try to figure out a more sensible name than the base64 # content of the data. origin = request.originatingObject() try: filename_url = origin.url() except AttributeError: # Raised either if origin is None or some object that doesn't # have its own url. We're probably fine with a default fallback # based on the data URL then. pass return urlutils.filename_from_url(filename_url, fallback='qutebrowser-download') def get_request(self, request, *, target=None, suggested_fn=None, **kwargs): """Start a download with a QNetworkRequest. Args: request: The QNetworkRequest to download. target: Where to save the download as downloads.DownloadTarget. suggested_fn: The filename to use for the file. **kwargs: Passed to _fetch_request. Return: The created DownloadItem. """ # WORKAROUND for Qt corrupting data loaded from cache: # https://bugreports.qt.io/browse/QTBUG-42757 request.setAttribute(QNetworkRequest.Attribute.CacheLoadControlAttribute, QNetworkRequest.CacheLoadControl.AlwaysNetwork) if suggested_fn is None: suggested_fn = self._get_suggested_filename(request) return self._fetch_request(request, target=target, suggested_filename=suggested_fn, **kwargs) def _fetch_request(self, request, *, qnam=None, **kwargs): """Download a QNetworkRequest to disk. Args: request: The QNetworkRequest to download. qnam: The QNetworkAccessManager to use. **kwargs: passed to fetch(). Return: The created DownloadItem. """ if qnam is None: qnam = self._networkmanager reply = qnam.get(request) return self.fetch(reply, **kwargs) @pyqtSlot('QNetworkReply') def fetch(self, reply, *, target=None, auto_remove=False, suggested_filename=None, prompt_download_directory=None): """Download a QNetworkReply to disk. Args: reply: The QNetworkReply to download. target: Where to save the download as downloads.DownloadTarget. auto_remove: Whether to remove the download even if downloads.remove_finished is set to -1. suggested_filename: The filename to use for the file. prompt_download_directory: Whether to prompt for a location to download the file to. Return: The created DownloadItem. """ if not suggested_filename: try: suggested_filename = target.suggested_filename() except downloads.NoFilenameError: _, suggested_filename = http.parse_content_disposition(reply) log.downloads.debug("fetch: {} -> {}".format(reply.url(), suggested_filename)) download = DownloadItem(reply, manager=self) self._init_item(download, auto_remove, suggested_filename) if download.cancel_for_origin(): return download if target is not None: download.set_target(target) return download # Neither filename nor fileobj were given filename = downloads.immediate_download_path(prompt_download_directory) if filename is not None: # User doesn't want to be asked, so just use the download_dir target = downloads.FileDownloadTarget(filename) download.set_target(target) return download # Ask the user for a filename question = downloads.get_filename_question( suggested_filename=suggested_filename, url=reply.url(), parent=self) self._init_filename_question(question, download) message.global_bridge.ask(question, blocking=False) return download def has_downloads_with_nam(self, nam): """Check if the DownloadManager has any downloads with the given QNAM. Args: nam: The QNetworkAccessManager to check. Return: A boolean. """ assert nam.adopted_downloads == 0 for download in self.downloads: assert isinstance(download, DownloadItem), download if download._uses_nam(nam): # pylint: disable=protected-access nam.adopt_download(download) return nam.adopted_downloads def init(): """Initialize the global QtNetwork download manager.""" download_manager = DownloadManager(parent=QApplication.instance()) objreg.register('qtnetwork-download-manager', download_manager) quitter.instance.shutting_down.connect(download_manager.shutdown)