Oops! Transport lines effectviely never appear aside from raw bridge
descriptors (which we never see), so I didn't have a live example to
test with.

Now we have one. DocTor's descriptor validation check is failing with...

  03/05/2020 00:35:33 [WARNING] Unable to retrieve the extrainfo descriptors: Transport line has a malformed address: transport obfs4 [2001:985:e77:5:fd34:f56b:c2d1:e98c]:10394 cert=dJ/a+vnP+eFv7FDaVUqWCVlyrqf8FlOva2YAEkDUwiGQuorZf4Oc6FXSdyn8b4pUmZj/WA,iat-mode=0

Caught thanks to GeKo.

Removing a private class that supported version rule chains (for example,
"check if I'm greater than 1.2.2 in the 1.2 series, or 1.3.5 in the 1.3
series"). This was only used for a single requirement check which has now
been removed.

We can always revive this class in the future if it's useful.

I doubt anyone has ever used event versions from our Requirement constants.
Retaining these within the event classes but dropping the clutter from our
version module.

Audited our library's cryptography imports for exception handling when
unavailable, and consistently document when it's required.

Python includes sqlite3 as a builtin, but when you compile python yourself it
isn't included by default. Our stem.manual attempted to account for this but
our tests didn't, and with python3 the module raises a ModuleNotFoundError
rather than an ImportError.

  Traceback (most recent call last):
    File "/home/atagar/Desktop/stem/test/task.py", line 160, in _import_tests
      importlib.import_module(module.rsplit('.', 1)[0])
    File "/home/atagar/Python-3.8.1/Lib/importlib/__init__.py", line 127, in import_module
      return _bootstrap._gcd_import(name[level:], package, level)
    File "<frozen importlib._bootstrap>", line 1014, in _gcd_import
    File "<frozen importlib._bootstrap>", line 991, in _find_and_load
    File "<frozen importlib._bootstrap>", line 975, in _find_and_load_unlocked
    File "<frozen importlib._bootstrap>", line 671, in _load_unlocked
    File "<frozen importlib._bootstrap_external>", line 783, in exec_module
    File "<frozen importlib._bootstrap>", line ...

When an Ed25519Certificate was created through from_base64() we appended an
undocumented 'encoded' attribute with the base64 content we were created from.
Our to_base64() method computes the exact same thing.

This attribute was only used in a single place for server descriptor
validation.

If 'encoded' and to_base64() mismatches that is a bug. Fiddling with these I
cound't come up with a scenario where that is a case, so dropping the redundant
attribute.

Also merging _validate_server_desc_signing_key() into its sole caller.

Our prereq module provides dependency checks, mostly for python versioning
(now moot) and the cryptography module.

It is not only cleaner but less error prone to perform dependency checks down
where their imports are performed rather than upfront. This way import changes
can't fall out of sync with our upfront checks.

Oops, when removing unused 'sign' arguments I accidently dropped our
skip_crypto_validation argument here. George and I invested quite a bit of
effort into creating cryptographically valid HSv3 descriptors, but HSv2
descriptors aren't.

This caused the following test failures...

  ======================================================================
  ERROR: test_minimal_hidden_service_descriptor
  ----------------------------------------------------------------------
  Traceback (most recent call last):
    File "/home/atagar/Desktop/stem/test/unit/descriptor/hidden_service_v2.py", line 414, in test_minimal_hidden_service_descriptor
      desc = HiddenServiceDescriptorV2.create()
    File "/home/atagar/Desktop/stem/stem/descriptor/hidden_service.py", line 702, in create
      return cls(cls.content(attr, exclude), validate = validate)
    File "/home/atagar/Desktop/stem/stem/descriptor/hidden_service.py", line 723, in __init__
      signed_digest = self._digest_for_signature(self.permanent_key, self.signature)
    File "/home/atagar/Desktop/stem/stem/descriptor/__init__.py", line 1043, in _digest_for_signature
      key = load_der_public_key(_bytes_for_block(signing_key), default_backend())
    File "/home/atagar/.local/lib/python3.7/site-packages/cryptography/hazmat/primitives/serialization/base.py", line 32, in load_der_public_key
      return backend.load_der_public_key(data)
    File "/home/atagar/.local/lib/python3.7/site-packages/cryptography/hazmat/backends/openssl/backend.py", line 1182, in load_der_public_key
      self._handle_key_loading_error()
    File "/home/atagar/.local/lib/python3.7/site-packages/cryptography/hazmat/backends/openssl/backend.py", line 1373, in _handle_key_loading_error
      raise ValueError("Could not deserialize key data.")
  ValueError: Could not deserialize key data.

Oops, I forgot to install cryptography under python 3.7 so I didn't exercise
several tests. This doesn't fix all the failures but corrects the most obvious
(we removed the sign argument when it was irrelevant).

Oops, I removed our LogBuffer in commit 76b6b49f but it's still used by
run_tests.py when we have a '--log' argument. Caught thanks to our Jenkins
tests...

  Traceback (most recent call last):
    File "./run_tests.py", line 478, in <module>
      main()
    File "./run_tests.py", line 238, in main
      handler = logging_buffer = stem.util.log.LogBuffer(args.logging_runlevel)
  AttributeError: module 'stem.util.log' has no attribute 'LogBuffer'

Doubtless there will be more opportunities to simplify, but for now cleaning
out the low hanging fruit. Dropping Sphinx documented deprecations and simple
TODO comments.

Earlier I replaced our usage of stem.util.conf as a cache with sqlite. Faster,
cleaner, and clearly the right option. We can now drop the legacy code.

Now that we're using python 3.x our sqlite module's uri mode can allow us to
open our cache in a read-only manner...

  https://docs.python.org/3/library/sqlite3.html#sqlite3.connect

Huh, odd. I could've sworn that I called this BaseHiddenServiceDescriptor
because HiddenServiceDescriptorV2 was originally called that so it still
retained an alias for the old name but I'm not spotting it. Oh well.

Renaming this class to more fittingly live aside their v3 counterpart.

Huh, wonder why this code was still around. Ticket 28615 was closed a long
while ago...

The pep8 module renamed itself to pycodestyle at Guido's request a while ago.
Stem was compatible with both but this was long enough ago that we can drop
support for the old module.

Just a few classes implemented a sign argument. Subclasses already customize
their signature when necessary so we can limit it to just the classes that use
it.

Couple attributes that were never used (they were misdocumentations in the
spec).

Redundant with the to_base64() method.

Old alias for v3ident.

This attribute was never implemented within tor.

Microdescriptor identifiers can appear multiple times so these attributes were
replaced with a hash.

Back in 2017 we changed this GETINFO option's behavior to return an empty list
rather than error when no results are available. THis is long enough that we
can drop this normalization.

  https://trac.torproject.org/projects/tor/ticket/21329

Odd, we have both a test_get_server_descriptor and test_get_server_descriptors
(plural) integration test but only the later required the online target. Here
at a coffee shop I got the following so adding this requirement to the other as
well.

  ======================================================================
  ERROR: test_get_server_descriptor
  ----------------------------------------------------------------------
  Traceback (most recent call last):
    File "/home/atagar/Desktop/stem/test/require.py", line 58, in wrapped
      return func(self, *args, **kwargs)
    File "/home/atagar/Desktop/stem/test/integ/control/controller.py", line 1301, in test_get_server_descriptor
      test_relay = self._get_router_status_entry(controller)
    File "/home/atagar/Desktop/stem/test/integ/control/controller.py", line 1489, in _get_router_status_entry
      for desc in controller.get_network_statuses():
    File "/home/atagar/Desktop/stem/stem/control.py", line 482, in wrapped
      for val in func(self, *args, **kwargs):
    File "/home/atagar/Desktop/stem/stem/control.py", line 1979, in get_network_statuses
      raise stem.DescriptorUnavailable('Descriptor information is unavailable, tor might still be downloading it')
  stem.DescriptorUnavailable: Descriptor information is unavailable, tor might still be downloading it

Removing the constants and version checks for obsolete tor versions. As of
2/6/20 our consensus has...

  client-versions 0.3.5.7,0.3.5.8,0.3.5.9,0.4.1.2-alpha,0.4.1.3-alpha,0.4.1.4-rc,0.4.1.5,0.4.1.6,0.4.1.7,0.4.1.8,0.4.2.1-alpha,0.4.2.2-alpha,0.4.2.3-alpha,0.4.2.4-rc,0.4.2.5,0.4.2.6,0.4.3.1-alpha
  server-versions 0.3.5.8,0.3.5.9,0.4.1.2-alpha,0.4.1.3-alpha,0.4.1.4-rc,0.4.1.5,0.4.1.6,0.4.1.7,0.4.1.8,0.4.2.1-alpha,0.4.2.2-alpha,0.4.2.3-alpha,0.4.2.4-rc,0.4.2.5,0.4.2.6,0.4.3.1-alpha

Stem won't have another release for several months so picked 0.3.5.7 as the
minimum version to keep. That narrows us down to HSv3 support for HSFETCH and
tor's new dormant mode.

We deprecated this attribute in favor of their parsed 'certificate'. Anyone
using 'my_descriptor.ed25519_certificate' can replace it with
'my_descriptor.certificate.encoded' to do the same.

Server descriptor files within tor's data directory begin with a couple
annotations. This is the only place where these annotations exist, and
is not part of tor's dir-spec.

I parsed these annotations for completeness when I first authored Stem.
However, we effectively never read these files nowadays for a few reasons...

  1. Stem's remote and collector modules provide better methods for getting
     descriptors.

  2. When tor is running locally we can get this information via GETINFO rather
     than reading directly from disk.

  3. Tor is trying to discourage us from reading the data directory from disk
     by restricting its permissions.

This isn't to say there aren't merits to reading on-disk descriptors. Nyx does
so, for instance, because it's a passive listener (no downloading allowed)
and massive GETINFO responses noticably lag the control port.

However, reading on-disk descriptors is rare and even in this edge case I'm
unaware of anyone using the annotations.

Redundant with the DirectoryAuthority's from_cache() function.

Removing a couple very old methods that are redundant with connect.

The consensus is relatively large, so parsing costs time and cpu. Doing this
when the parsed content is requested rather than when the event is first
received.

Nyx disabled eager parsing long ago because otherwise this is a significant
periodic performance hit.

A simple string-to-string hash could not convey multi-value configuration
values so we replaced it.

This attribute was a moving target that frequenly got out of date so we
deprecated it a while ago.

Removing get_socks_listeners and is_geoip_unavailable as both have better
counterparts nowadays.