Private windows are opened as normal tabs after a crash

[chronus7]

Description

After dismissing the bug/crashreport-dialog of qutebrowser after a crash, qutebrowser opens a window with all previous tabs in it. This includes tabs from previously private windows.

Expected Behaviour

Any of the three:

1. One single window with only the startpage (like Control-n).
2. All previous windows (possibly except the private ones) with their respective tabs.
3. The latest version of the default-session.

Version

This behaviour existed already before the current version.

qutebrowser v1.3.1
Git commit: 
Backend: QtWebEngine (Chromium 65.0.3325.151)

CPython: 3.6.5
Qt: 5.11.0
PyQt: 5.10.1

sip: 4.19.8
colorama: 0.3.9
pypeg2: 2.15
jinja2: 2.10
pygments: 2.2.0
yaml: 3.12
cssutils: no
attr: 18.1.0
PyQt5.QtWebEngineWidgets: yes
PyQt5.QtWebKitWidgets: no
pdf.js: no
sqlite: 3.23.1
QtNetwork SSL: OpenSSL 1.1.0h  27 Mar 2018

Style: QFusionStyle
Platform: Linux-4.16.13-1-ARCH-x86_64-with-arch, 64bit
Linux distribution: Arch Linux (arch)
Frozen: False
Imported from /usr/lib/python3.6/site-packages/qutebrowser
Qt library executable path: /usr/lib/qt/libexec, data path: /usr/share/qt

Paths:
cache: $HOME/.cache/qutebrowser
config: $HOME/.config/qutebrowser
data: $HOME/.local/share/qutebrowser
runtime: /run/user/1000/qutebrowser
system data: /usr/share/qutebrowser

[The-Compiler]

This is because qutebrowser simply passes open pages as an argument to qutebrowser when reopening. Originally this was before sessions were implemented, but I kept it that way because there's a smaller chance (because of less complexity) that something would go wrong when restoring the pages.

However, this definitely shouldn't re-open formerly private tabs as normal ones. I'm tempted to just use sessions for crash recovery anyways, because at that point it's probably easier to do so (and it has some additional benefits).

[user202729]

Won't using sessions have the disadvantage of not being able to be "forgiving"? Or are you intending to implement a forgiving mode for session-save?

---

Given that currently the autosave session is always immediately saved on page load (at least until PR 4087 is merged), it should be easier to do nothing (on start, qutebrowser already loads the autosave file if there's one)

[The-Compiler]

What do you mean with "forgiving"?

[user202729]

What do you mean with "forgiving"?

qutebrowser/qutebrowser/misc/crashsignal.py

Line 121 in 34be0a8

def _recover_pages(self, forgiving=False):

[The-Compiler]

Would probably make sense to have something like that for :session-save (or rather, the underlying code) then, yeah.