Back in our old 'arm' days users could configure us through a file at ~/.armrc.
Nyx changed this location to ~/.nyx/config but continued to call it a 'nyxrc'.
Caught thanks to Sai...

  https://github.com/torproject/nyx/pull/27

Mostly this changes internal references, but there are two user visible
changes...

  * This fixes our FAQ which incorrectly said that we read from ~/.nyx/nyxrc.

  * Removes the sample nyxrc path from our man page. Our setup.py no longer
    installs such a file.

This intentionally keeps two references to a nyxrc...

  * Our website's url for a sample is...

      https://nyx.torproject.org/nyxrc.sample

    We can't change this prior to our next release without confusing users.

  * Our legacy changelog entries still call it a nyxrc because that's what it
    was called back then.

I don't run Nyx's unit tests very often so a few regressions slipped in. Some
due to Nyx adjustments, some from Stem.

Stem renamed its is_pep8_available() function to
is_pycodestyle_available() because the upstream project renamed itself...

  Traceback (most recent call last):
    File "./run_tests.py", line 80, in <module>
      main()
    File "/home/atagar/Desktop/nyx/stem/util/conf.py", line 289, in wrapped
      return func(*args, **kwargs)
    File "./run_tests.py", line 46, in main
      if stem.util.test_tools.is_pep8_available():
  AttributeError: module 'stem.util.test_tools' has no attribute 'is_pep8_available'

There was a permission issue. Trying again.

Pushing a commit to hopefully trigger this repository to be mirrored within
GitHub (https://github.com/torproject/nyx).

Oops, my recent change to skip the installation test when offline broke
python3...

  ======================================================================
  ERROR: installation (unittest.loader._FailedTest)
  ----------------------------------------------------------------------
  ImportError: Failed to import test module: installation
  Traceback (most recent call last):
    File "/usr/lib/python3.5/unittest/loader.py", line 428, in _find_test_path
      module = self._get_module_from_name(name)
    File "/usr/lib/python3.5/unittest/loader.py", line 369, in _get_module_from_name
      __import__(name)
    File "/home/atagar/Desktop/nyx/test/installation.py", line 6, in <module>
      import urllib2
  ImportError: No module named 'urllib2'

On my morning bus commute I was pretty puzzled why our tests broke. Turns out
we fail in a rather confusing way that says our PYTHONPATH is misconfigured
when unable to install stem from pypi.

Oops, line wrapping broke our tests...

  ======================================================================
  FAIL: test_multiline_panel (panel.interpreter.TestInterpreter)
  ----------------------------------------------------------------------
  Traceback (most recent call last):
    File "/home/atagar/Desktop/nyx/test/__init__.py", line 59, in wrapped
      return func(self, *args, **kwargs)
    File "/usr/local/lib/python2.7/dist-packages/mock/mock.py", line 1305, in patched
      return func(*args, **keywargs)
    File "/home/atagar/Desktop/nyx/test/panel/interpreter.py", line 107, in test_multiline_panel
      self.assertEqual(EXPECTED_MULTILINE_PANEL, test.render(panel._draw).content)
  AssertionError: 'Control Interpreter:\n>>> GETINFO version\n250-version=0.2.4.27 (git-412e3f7dc9c6c01a)\n>>> to use this panel press enter' != u'Control Interpreter:\n>>> to use this panel press enter'

  ======================================================================
  FAIL: test_scrollbar (panel.interpreter.TestInterpreter)
  ----------------------------------------------------------------------
  Traceback (most recent call last):
    File "/home/atagar/Desktop/nyx/test/__init__.py", line 59, in wrapped
      return func(self, *args, **kwargs)
    File "/usr/local/lib/python2.7/dist-packages/mock/mock.py", line 1305, in patched
      return func(*args, **keywargs)
    File "/home/atagar/Desktop/nyx/test/panel/interpreter.py", line 120, in test_scrollbar
      self.assertEqual(EXPECTED_WITH_SCROLLBAR, test.render(panel._draw).content)
  AssertionError: 'Control Interpreter:\n |>>> GETINFO version\n |250-version=0.2.4.27 (git-412e3f7dc9c6c01a)\n |\n |\n |\n |\n |\n |\n |\n |\n |\n |\n |\n |\n |\n |\n |\n |\n |\n |\n |\n |\n |\n-+' != u'Control Interpreter:\n>>> to use this panel press enter'

  ----------------------------------------------------------------------

With commit 75052e8d tests should no longer directly modify their private
'_lines' attribute. We need to call the _add_line() helper instead to ensure
the wrapped and unwrapped attributes stay in sync.

Oops, commit 54532316 changed our argument parsing behavior in a way that broke
our tests.

Just a couple minor regressions from prior changes.

Our installation unit test results in a couple egg things. Cleaning those up
afterward.

Adjustments to our setup.py broke the installation test. Got really sure why it
started causing errors about pth but easy enough to work around.

When we bump our version our about dialog rightfully changes. Revising our test
not to expect a hardcoded version.

Got additional testers on #tor-dev and turns out our custom installer causes it
to fail for just about everybody. Dropping the ability to specify a man or
samplerc path to get us back on the road best traveled. Also swapping from
distutil to setuptools so we have entry_points (which we need for a bin
script).

Couple recent changes need to be accounted for.

Now that stem can cache our address we can drop our our_address() helper.

Lots of breakages have slipped in while attempting to improve our performance.
Most were timing related due to our switch to static sleeps.

At only INFO runlevel cpu usage on my relay jumps from 13% to 20%. This is
entirely coming from deduplication. We made this far more efficient (linear
from n^2) but it's pretty easy to finish dropping this down to a constant
time operation.

Our graph panel updates and redraws itself every second, making it a
non-trivial load on cpu. Two actions in particular account for most of
it...

  * Cloning stats on every redraw. This constantly consumes 0.3% cpu on my
    relay. Being lock-free is nice, but in this case a lock really makes
    sense.

  * Performing a flurry of addstr calls for each cell in the graph. Doing
    a single vline call for each column instead drops cpu by another 0.4%.

On my relay this drops cpu usage of the graph panel from 1% to 0.4% (60%
less)...

  * 0.3% cpu usage when only the header panel is rendered
  * 1.3% cpu usage when the graph panel was rendered too
  * 0.7% cpu usage when the graph panel was rendered with these changes

Just adding some basic coverage for the rest of the argument module.

There's a few things to consult to determine our chroot so making this a
helper rather than part of the starter.

When interface is set we unset the socket, and vic versa. Doing this in the
arguments module both makes sense and lets us clean up the starter a bit.

If it's been less than an hour there's no point in making this GETINFO call.
Hopefully this will speed startup a bit.

There, that's more like what I was after. Adding methods to read and write
relay data to the cache along with tests.

Well, that didn't last very long. Got annoyed by calling 'nyx.cache.cache()'.
Oh well.

Now that we have a class that can do its own locking there's no need to require
a context manager.

Not yet even close to what I'm planning, but moving this out of the base module
is a good first step.

Our initial use for the cache is pretty clear so making a table for it and
adjusting our tests to use it.

Versioning nyx's cache schema and ensure use of the cache is done in a thread
safe fashion.

Simple helper that provides back a sqlite cache for us to use. If we can
persist it then great, if not then it resides in memory.

Fetching just the manual information we need, when we need it. Rather than
loading the whole manual this now retrieves information from the sqlite manual
cache. This should mean slightly faster startup and reduced memory usage.
However, this no longer parses the local man page (so information cannot be
more up-to-date than stem's cache).

We can re-introduce usage of local man pages later if this becomes something
folks care about.

Our date bounding box was placed in the completely wrong spot (inverted x/y
coordinates), and the side border vanished in the top wasn't visible.

Both our header and graph frequenly refer to our router status entry. In
practice this only changes once per hour so caching it for five minutes. At
debug runlevel this causes the vast majority of the noise in our logs.

Oops! Could've sworn I ran our tests after our earlier changes but evidently
not. Moving our nyxrc.sample broke installation, and swapping to bytes broke
some test assertions.

Damn the scrollbars have been a bigger headache than anything else thus far.
There was an off-by-one error where we compared the number of scroll elements
to the last index.

Newer version of pycodestyle does some additional validation. Some aren't
issues we're concerned with.

Damn this scrollbar method has given me a lot of headaches. Correcting an
off-by-one error that caused the tests to fail on my netbook but pass on my
pc.

I really hope this is the last time I need to weed one of these out. For my
personal reference here's the numbers when working through the test...

  top = 15
  top_index = 21
  size = 30
  self.height = 25

  scrollbar_height = self.height - top - 1 = 25 - 15 - 1 = 9

  bottom_index = top_index + scrollbar_height = 21 + 9 = 30

  slider_top = (scrollbar_height * top_index) // size = (9 * 21) // 30 = 6
  slider_size = (scrollbar_height * (bottom_index - top_index + 1)) // size = (9 * (30 - 21 + 1)) // 30 = 3