Adding a check that the authentication cookie is 32 bytes before sending its
contents to the control port. This is to prevent a malicious socket from
tricking us into reading them arbitrary file content. Tested by hardcoding an
alternative file as being the cookie and confirming that this makes arm abort
initialization. Caught by rransom.

https://trac.torproject.org/projects/tor/ticket/4305

In new tor versions the DirReqStatistics is enabled by default and disabled
when geoip lookups are unavailable. This casuses arm to warn that the torrc
differs from tor's state. Discarding this validation warning instead. Tested by
checking that validation didn't complain when I SETCONF this option. Caught by
Sebastian.

https://trac.torproject.org/projects/tor/ticket/4237

BSD platforms lack either pwdx or proc contents with this information. However,
this is available via lsof. Using this for BSD platforms. This was tested by
running sysTools.getPwd() on OSX - hopefully it should work on Free/OpenBSD
too. Fix is thanks to Sebastian.

https://trac.torproject.org/projects/tor/ticket/4236

Caught by Sebastian:
https://trac.torproject.org/projects/tor/ticket/4237

Linux resolvers naturally fail on OpenBSD. I'm not sure if the bsd variants of
procstat/sockstat are installable (they're definitely not around by default) so
falling back ot only lsof like macs. Thanks to a test system from Jordi Espasa
Clofent.

Linux and BSD platforms (mac, openbsd, and freebsd) have distinct argument sets
for their ps resolvers. Taking into account for this for tor detection,
guessing the family based on uname results then trying both if we guess wrong.

Adding the '-a' argument to get results for all users thanks to Jordi Espasa
Clofent.

Tested this ps command on Mac and OpenBSD, and just guessing that FreeBSD will
behave the same.

In order to use the fresh values a fromCache argument is passed.

OpenBSD uses a different variant of ps causing the tor process detection (among
many other things) to fail. Caught thanks to a test system from Jordi Clofent.

Conflicts:
	src/util/__init__.py

Evidently it isn't enough for panels to clear their contents, so also
requesting a clear of the uppermost stdscr. Fix suggested by np...
https://trac.torproject.org/projects/tor/ticket/2830#comment:9

It makes more sense for the deb flag to remove packages from the base install
rather than the omission adding. The rpm builds will be doing something similar
so making the tweak now so we can add the rpm flag to the torctl removal check.

The arm gui is still a prototype and shouldn't introduce requrements for end
users, or even be packaged yet. This properly does dependency checks when the
user *requsts* the gui (fetching cagraph if unavailable) so this should make
everyone happy. :)

The ctrl+L keybinding is suppose to perform a redraw but this was only making a
'request' rather than forcing it, so unchanged portions of the interface
weren't actually redrawn. Caught by np.

https://trac.torproject.org/projects/tor/ticket/2830

When the user presses 'esc' in the interpretor panel prompt the getstr result
is None. I didn't account for this, causing a stacktrace. Caught by Sebastian.

https://trac.torproject.org/projects/tor/ticket/4098

Tom Lowenthal authored 13 years ago and Damian Johnson committed 13 years ago

Added log duplicate masks for longer-than-expected circuit build times and re-estimates of slow circuits.

Forgot to fill in the trac urls for the tickets when making the changelog for
this last release.

Renaming the prompt function and format variable to avoid a naming conflict
with other uses of that name in the module. Not actually an error, but could
lead to headaches if it did cause a bug. Caught by pylint.

Notice level message at startup that cautions the user against running arm as
root.

The getRelayAddress function would ignore its default argument if tor is shut
down, causing it to return None and make arm crash. This is probably a timing
issue which is why it went undiscovered so long.

Notice level message at startup that cautions the user against running tor as
root.

When arm failed to query tor's connections it would give a short warning saying
so. This often spooked users so lowering this to be a NOTICE and giving more
information on both what happened and how to fix it (99% of the time this is a
simple permissions issue). Change requested by Sebastian

The startup wizard option is currently unimplemented and disabled in the
setting.cfg. However, there was an os compatability check left over from when I
was gonna implement it. That and its comment were confusing since the option's
not implemented so commenting it out with a more up to date expanation.

When the user opts for the generated torrc to be for the system instance the
wizard shows the wrong path in the confirmation panel.

Users could cancel the wizard by prssing esc, but it's also intuitive to expect
'q' to do the same (since that's the keybinding to quit arm entirely).
Respecting this option too. This was requested by monochromec on...
https://trac.torproject.org/projects/tor/ticket/3995

This also fixes a minor bug where the config options dialog wouldn't respect a
cancel signal (instead looping on that dialog).

If we got a CIRC event while populating the cache of relays that we're attached
to we could reset _fingerprintsAttachedCache to None while we were in the
middle of populating it. Manipulation of this cache is supposed to be governed
by the connection lock, so fixing that for the CIRC event's reset.

Occasionally people are confused by the notice that no armrc was loaded,
thinking that it's a problem. Trying to clarify the message (it's just supposed
to inform users that they can customize arm's behavior).

This is a hacked up version of the fix for
https://trac.torproject.org/projects/tor/ticket/3958

It's a bit of a Frankenstein and will be removed when the TorCtl fix is merged.

When checking to see if a tor process is already running I forgot the '-A' flag
for ps, limiting the check to our user.