pynacl-1.0.1 has a bug, on systems which have the sodium runtime library
installed (/usr/lib/libsodium.so, as provided by a package like
"libsodium13"), but not the development headers (/usr/include/sodium.h,
as provided by "libsodium-dev"). The pynacl setup.py tries to detect
whether a system libsodium can be used (instead of the bundled copy), by
using cffi to load libsodium.so and check the version strings. However
this test doesn't check that sodium.h is available, which is needed to
build the glue code.
The long-term fix is to change pynacl, either to improve the test (to
look for a header file, which sounds tricky), or to always use the
bundled sodium unless explicitly told otherwise (removing the test
altogether). The short term fix is to tell magic-wormhole builders to
install libsodium-dev if they run into this error, or use the
environment-variable override when installing.
closes#10
refs https://github.com/pyca/pynacl/issues/184
These weren't running because Click complained about an ASCII locale
when running under py3, which triggered an error check that was there to
detect broken virtualenvs, skipping those tests.
The fix appears to be to force the en_US.UTF-8 locale when running the
wormhole program in a subprocess.
This adds a test for database upgrades, which I developed on a branch
that added a new DB schema (v3) and an upgrader to match, but then I
changed my mind about the schema and removed that part. The test will be
useful some time in the future when I change the schema in a small
enough way that I bother to write an upgrader for the change. For now,
the test is disabled.
In addition, the upgrader test is kind of lame. I'd really prefer to
assert that the upgraded schema is identical to the schema of a
brand-new (latest-version) database, but ALTER TABLE doesn't quite work
that way (comments are omitted, and the order of the columns is slightly
different).
This also adds database.dump_db() for the tests.
There was some vestigal server-cli code (leftover in the client-side
wormhole.cli.cli_args) that used port 3000/3001, and it accidentally got
used for the new Click-based parser, rather than the actual server-cli
code (in wormhole.server.cli_args) that uses port 4000/4001. This
changes the port numbers to match (everything uses 4000/4001 these days,
to avoid confusing interactions with the old 0.7.6 server that might
still be listening on the old ports).
GNU libreadline, and the libedit-based library shipped on stock OS-X
python, require different key-binding syntaxes to enable tab completion.
The previous commit to fix this (0977ef0) added both binding commands
Unfortunately when GNU libreadline is given the libedit-style
command (i.e. "bind ^I rl_complete"), it binds the letter "b" to a
non-existent command "ind", or something, and as a result the letter "b"
doesn't work anymore.
This patch uses the readline docstring to sense which flavor is
installed, and only runs the one binding command that's appropriate.
refs #37
The appveyor tests were failing because their VMs only have 127.0.0.1,
and stripping it out resulted in an empty hint list, which meant Transit
couldn't work at all.
