Jeremy Stanley [Sat, 16 Sep 2017 17:35:48 +0000 (17:35 +0000)]
Data nodes starting with "_" are metadata
Avoid loading data nodes with a "_" prefix as Universe content.
These are only used as file-local metadata.
Jeremy Stanley [Sat, 16 Sep 2017 17:29:38 +0000 (17:29 +0000)]
Interpret a "." prefix as the startdir
Since daemonization explicitly changes the working directory to "/",
specifying a mudpy.filing.prefix of "." should be taken to refer to
the original starting directory recorded in Universe.startdir rather
than the current directory for the process.
Jeremy Stanley [Sun, 27 Aug 2017 17:07:45 +0000 (17:07 +0000)]
Overhaul data management to get rid of __control__
Remap __control__ subkeys in data as follows:
default_files -> .mudpy.filing.categories
include_dirs -> _load
include_files -> _load
read_only -> _lock
Additionally, obsolete the __control__.private_files key by
switching from raw file/directory names to associative arrays in the
category defaults so that flags such as "private" can be added to
them, and make the category filenames automatically determined so
that they can be omitted unless specific overrides are required.
For the sake of future-proofing, rename the DataFile class to Data
and its "filename" attribute to "source" so that we avoid confusing
renames later when these might instead refer to some other storage
medium such as a table in a database. Similarly rename the Element
class attribute "filename" to "origin" attribute both to make it
more generic and to avoid confusion with "source" (an Element's
origin is a complete Data object, while a Data's source is just a
rooted-anchored file path currently).
Add a Universe.add_category convenience method to handle properly
guessing the corresponding fallback path and copying any declared
flags.
Jeremy Stanley [Mon, 31 Jul 2017 15:14:08 +0000 (15:14 +0000)]
Error if set command is used on read-only element
It is possible for an admin to accidentally attempt to set a facet
on an element from a read-only file. If this happens, trap the
PermissionError exception and return a clear error message to the
user instead of crashing the service. Also add a regression test to
make sure this continues to work as intended, since preventing
alteration of read-only elements is a critical security measure we
need to make certain we preserve going forward.
Jeremy Stanley [Mon, 31 Jul 2017 15:00:33 +0000 (15:00 +0000)]
Refuse to alter read-only elements at runtime
In the Element.set() method, check whether the element being altered
is from a read-only file. That should only ever happen when
(re)loading of files is underway, either at start time or because
the reload command/SIGHUP has been issued. If an attempt is made to
change a read-only element at any other time, raise a
PermissionError exception instead of setting the new value in memory
(altered values were never written to the backing file, but prior to
this it was possible to modify the in-memory copies).
Jeremy Stanley [Mon, 31 Jul 2017 14:56:57 +0000 (14:56 +0000)]
Track whether loading is underway
In preparation for safeguards preventing alteration of elements from
read-only backing files, we need a way to identify that loading
updated files is in progress; that's the only time elements from
read-only files should ever change.
Jeremy Stanley [Mon, 31 Jul 2017 14:54:34 +0000 (14:54 +0000)]
Test the set command
Add a test to make sure the set command works, and use it to alter
the facet we're subsequently checking in the writeable element for
the show element test to make sure it actually worked.
Jeremy Stanley [Mon, 31 Jul 2017 14:49:09 +0000 (14:49 +0000)]
Use writeable element in show element test
In preparation for being able to test the set command, switch the
legacy element being shown to one which is not in a read-only file
so that we'll be able to test altering it in a future change.
Jeremy Stanley [Sun, 30 Jul 2017 15:52:46 +0000 (15:52 +0000)]
Update copyright dates for files changed this year
Bump the copyright date to the present for files which received
significant modification since the beginning of the year.
Jeremy Stanley [Sun, 30 Jul 2017 02:40:25 +0000 (02:40 +0000)]
Rename internal:storage to .mudpy.filing
Convert the old style internal:storage element to a new style
.mudpy.filing element and adjust the find_file() function
accordingly.
Jeremy Stanley [Sat, 29 Jul 2017 18:42:30 +0000 (18:42 +0000)]
Stop unnecessarily stripping quote marks
Since the switch to YAML, quotations around strings are
automatically interpreted an so not included in the parsed result.
Remove legacy code which stripped leading and trailing quotes from
string values as this is no longer needed.
Jeremy Stanley [Sat, 29 Jul 2017 18:34:55 +0000 (18:34 +0000)]
Explicitly configure the root_path in examples
In the example and test configurations, set the root_path value to
"." signifying the current/starting directory. This exercises the
option so we can make sure it doesn't break anything.
Jeremy Stanley [Sat, 29 Jul 2017 18:31:40 +0000 (18:31 +0000)]
Log missing file inclusions as more severe
When another file is included with an inc macro in text, being
unable to load that file could indicate a fairly serious
configuration error. Increase the severity of the log message for
this so that it will be more easily noticed and caught in testing.
Jeremy Stanley [Sat, 29 Jul 2017 18:24:06 +0000 (18:24 +0000)]
Fix file pathing in daemon mode
If run as a daemon, the working directory is reset to "/" for
safety. Make sure that relative file searching takes this into
account and prefixes with the original (pre-chdir) start directory
rather than the current working directory.
Jeremy Stanley [Sat, 29 Jul 2017 16:18:16 +0000 (16:18 +0000)]
Rename internal:time to .mudpy.timing
Convert the old style internal:time element to a new style
.mudpy.timing element and adjust the User.check_idle() method and
on_pulse() function accordingly.
Jeremy Stanley [Sat, 29 Jul 2017 15:27:22 +0000 (15:27 +0000)]
Switch example listening port from 6669 to 4000
As a nod to the legacy of DikuMUD, use 4000/tcp as the example
listening socket instead of 6669/tcp.
Jeremy Stanley [Sat, 29 Jul 2017 15:05:26 +0000 (15:05 +0000)]
Fix facets method to give new-style relative names
For new-style elements correct the facets method to return short
(relative) key names, maintaining parity with old-style elements.
Add a new-style element to the "show element" selftest so that we
can make sure it won't regress.
Jeremy Stanley [Sat, 29 Jul 2017 15:01:08 +0000 (15:01 +0000)]
Use importlib in place of imp
As of Python 3.4, the newer importlib module supports the reload()
function we need. Switch from the deprecated imp module, except
before 3.4 where we will simply pretend imp is importlib for
readability.
Jeremy Stanley [Sat, 29 Jul 2017 14:53:36 +0000 (14:53 +0000)]
Reduce the log message priority for reload calls
It's important to log when the "reload" command is issued, but it's
not an error condition. Log the event at priority 6 rather than 8.
Jeremy Stanley [Sat, 29 Jul 2017 14:51:37 +0000 (14:51 +0000)]
Exercise the reload command in tests
To make sure it doesn't get broken in a coming change, call the
admin command `reload` in the selftest.
Jeremy Stanley [Sat, 29 Jul 2017 03:56:25 +0000 (03:56 +0000)]
Switch quoting styles to reduce escaped quotes
In an effort to address leaning toothpick syndrome, adjust quoting
styles where possible to avoid escaping quote marks without
increasing ambiguity.
Jeremy Stanley [Fri, 28 Jul 2017 20:03:58 +0000 (20:03 +0000)]
Use raw strings when escape sequences are needed
Python 3.6 has added a DeprecationWarning for any regular strings
containing non-doubled backslash escapes. Use raw strings for these
instead so we'll be ready when the old syntax is eventually
disallowed.
Jeremy Stanley [Sun, 2 Jul 2017 15:13:59 +0000 (15:13 +0000)]
Correct name of the custom loglevel test
Make the name of the custom loglevel test distinct from the show log
test preceeding it.
Jeremy Stanley [Sun, 2 Jul 2017 15:11:17 +0000 (15:11 +0000)]
Handle ValueError when setting incorrect type
Trap for ValueError exceptions when attempting to pass a value to
the set command with an incorrect data type for the facet being set,
returning a clear failure message to the calling user.
Jeremy Stanley [Thu, 15 Jun 2017 15:26:10 +0000 (15:26 +0000)]
Coerce loglevel facet to int when updating
Calling the set command always results in updating values to type
str, which can result in failures when those values are later
re-read. Explicitly coerce the loglevel value to int whenever it is
updated to avoid a subsequent exception when calling the show log
command.
This points out a need for maintaining a schema so that values can
be coerced to the appropriate type when needed, but for now the
conditional adjustment can serve as a placeholder for a more
thorough design.
Jeremy Stanley [Sun, 14 May 2017 18:23:24 +0000 (18:23 +0000)]
Clear vestigial ord() calls in telopt negotiation
Clean up missed ordinal recasting on telnet options during
negotiation. They're no longer necessary, and crash the service if
triggered.
Jeremy Stanley [Sun, 14 May 2017 17:42:40 +0000 (17:42 +0000)]
Stop using vary_rounds with passlib
The passlib library has deprecated the vary_rounds KDF parameter
since it's basically pointless as long as a strong salt is employed.
Proactively remove the reference as it will no longer be accepted by
passlib 2.0.
Jeremy Stanley [Sun, 14 May 2017 16:42:55 +0000 (16:42 +0000)]
Guard against KeyError on universe.contents index
When accessing base level configuration elements from the main
engine, check that they're actually in universe.contents before
trying to retrieve their facets. Also use sane default fallbacks if
they're undefined.
Jeremy Stanley [Sun, 14 May 2017 14:00:11 +0000 (14:00 +0000)]
Enable initial admin account in example config
The example etc/mudpy.yaml, intended for demonstration purposes,
only binds a listening socket on the loopback interface so shouldn't
be exposed remotely. As a result, turn on the configuration option
to enable elevating the account named "admin" to administrative
privileges automatically upon creation. This allows to run tests of
admin-specific commands with the example configuration.
Also rename the "testadmin" account used by the selftest to match
the name of the default for consistency, and rename test_config.yaml
to test_daemon.yaml reflecting its primary purpose (and also to open
the way for additional test configs as well).
Jeremy Stanley [Thu, 4 May 2017 19:34:08 +0000 (19:34 +0000)]
Ignore unpacked *.egg directories
When performing in-tree sdist installs, stray *.egg directories may
be left behind. Don't track them.
Jeremy Stanley [Mon, 17 Apr 2017 16:46:27 +0000 (16:46 +0000)]
Modernize coder documentation
Update the coder docs to remove obsolete references to the old
git2gch script, include new information on testing expectations and
outline a recommended tox-based developer environment and workflow.
Jeremy Stanley [Mon, 17 Apr 2017 16:03:57 +0000 (16:03 +0000)]
Add a generic tox env for sdist generation et al
Add a testenv:venv environment to the tox.ini as a convenience for
things like sdist/wheel building or generating documentation. Also
add the autogenerated files from running setup.py sdist and
bdist_wheel commands to the .gitignore file.
Jeremy Stanley [Tue, 28 Mar 2017 02:21:39 +0000 (02:21 +0000)]
Add a suitable test config for the daemon
In service of more repeatable testing and improved feature coverage,
include a test fixture for a version of the normal sample
configuration with some additional non-default features enabled and
some settings tuned to values which make testing easier/faster (not
recommended for production use):
* allow "testuser" to automatically be an admin when created
* enable data backup copies but keep only 3 when rotating
* turn on file-backed logging
* background the daemon process once initialized
* write a pidfile
* inject status updates into logs every 25 ticks
* save and rotate datafiles every 5 ticks
Jeremy Stanley [Mon, 20 Feb 2017 21:42:08 +0000 (21:42 +0000)]
Package the self-test script as an entrypoint
Move the self-test script into the Python package and define an
entrypoint for it in setup.cfg so a `mudpy_selftest` executable will
be installed.
Jeremy Stanley [Mon, 20 Feb 2017 06:32:53 +0000 (06:32 +0000)]
Correct filesystem permissions on daemon module
When the executable script was moved to become an in-package module,
its permissions were incorrectly left executable. Correct them to
match the other modules in the package.
Jeremy Stanley [Mon, 2 Jan 2017 06:06:41 +0000 (06:06 +0000)]
Make the self-test script importable
In preparation for moving the self-test script to a more standard
entrypoint, make it safely importable and give it a separate
conditional main() execution.
Jeremy Stanley [Fri, 16 Dec 2016 09:55:47 +0000 (09:55 +0000)]
Add missing copyright/license notices
For clarity, add copyright and license notice headers at the tops of
files which were missing them (general packaging configuration and
metadata for the most part, but doesn't hurt to be thorough for the
sake of future provenance inquiries).
Jeremy Stanley [Thu, 15 Dec 2016 11:25:38 +0000 (11:25 +0000)]
Support running flake8 under tox
Add the necessary configuration to be able to run `tox` for
performing style checks easily with flake8.
Jeremy Stanley [Thu, 15 Dec 2016 06:17:09 +0000 (06:17 +0000)]
Package the daemon executable as an entrypoint
Move the daemon executable into the Python package and define an
entrypoint for it in setup.cfg so a similar `mudpy` executable will
still be installed. Drop the unnecessary import path override now
while at it.
Jeremy Stanley [Wed, 14 Dec 2016 19:45:30 +0000 (19:45 +0000)]
Make the executable importable
In preparation for moving the executable to an more standard
entrypoint, make it safely importable and give it a separate
conditional main() execution.
Jeremy Stanley [Wed, 14 Dec 2016 01:05:09 +0000 (01:05 +0000)]
Create a Python package
Add setup.cfg and setup.py files sufficient to create a "mudpy"
Python package suitable for `pip install` and uploading to PyPI.
Jeremy Stanley [Thu, 24 Nov 2016 06:28:14 +0000 (06:28 +0000)]
Appease style checks
Once more the PEP8 gods stir to anger. Sate their fury with a
generous offering of whitespace, such that they might return to
their baleful slumber.
Jeremy Stanley [Wed, 23 Nov 2016 14:03:37 +0000 (14:03 +0000)]
Clean up trailing whitespace in documentation
Remove some unsightly trailing whitespace in the readme and coder
reference.
Jeremy Stanley [Thu, 20 Oct 2016 17:11:30 +0000 (17:11 +0000)]
Move documentation files for packaging preparation
Move doc/coder.txt into doc/source/ where Sphinx will expect it,
doc/LICENSE to the top-level directory renamed to LICENSE.rst for
clarity on its format, and similarly rename README to README.rst as
well.
Jeremy Stanley [Wed, 5 Oct 2016 10:02:35 +0000 (10:02 +0000)]
Move lib/mudpy to mudpy for packaging preparation
Move the lib/mudpy tree to just mudpy where it will ultimately
reside once integrated into a Python package. Temporarily adjust the
module search path addition in the executable until packaging work
is complete (at which point it can safely be dropped).
Jeremy Stanley [Sun, 18 Sep 2016 06:25:31 +0000 (06:25 +0000)]
Rename internal:process to .mudpy.process
Convert the old style internal:process element to a new style
.mudpy.process element and adjust the daemonize(), create_pidfile()
and remove_pidfile() functions accordingly.
Jeremy Stanley [Thu, 15 Sep 2016 06:44:48 +0000 (06:44 +0000)]
Rename internal:network to .mudpy.network
Convert the old style internal:network element to a new style
.mudpy.network element and adjust the initialize_server_socket()
function accordingly.
Jeremy Stanley [Wed, 24 Aug 2016 01:17:08 +0000 (01:17 +0000)]
Rename internal:directions to .mudpy.movement.*
Convert the old style internal:directions element to a series of new
style .mudpy.movement.* elements and adjust the move_direction()
function accordingly. Also extend the Universe class to provide a
set-based index of direction names found in the data.
Jeremy Stanley [Sun, 24 Jul 2016 01:14:10 +0000 (01:14 +0000)]
Rename internal:logging: to .mudpy.log
Convert the old style internal:logging element to a new style
.mudpy.log element and adjust the log() function accordingly.
Jeremy Stanley [Sun, 24 Jul 2016 00:45:36 +0000 (00:45 +0000)]
Update README for recent development
Edits to the README file reflect the switch to Py3K, YAML and
addition of non-stdlib dependencies.
Jeremy Stanley [Sun, 17 Jul 2016 02:45:00 +0000 (02:45 +0000)]
Rename internal:limits element to .mudpy.limit
Convert the old style internal:limits element to a new style
.mudpy.limit element and rename the default_admins,
default_backup_count and max_avatars facets to admins, backups and
avatars respectively.
Jeremy Stanley [Tue, 28 Jun 2016 07:46:08 +0000 (07:46 +0000)]
Strip extra separators in show element command
Eat leading and trailing "." separators in the show element command,
so that it will consistently search the element set for the
corresponding simplified element names.
Jeremy Stanley [Tue, 28 Jun 2016 06:50:50 +0000 (06:50 +0000)]
Fix 8-bit Telnet with Py3K
Replace the lambda filter for non-ASCII characters with a list
comprehension, correcting and simplifying the range comparison and
recombination of the filtered string.
Jeremy Stanley [Mon, 9 May 2016 05:09:14 +0000 (05:09 +0000)]
Rename internal:language: to .mudpy.linguistic
Convert the old style internal:language element to a new style
.mudpy.linguistic element and adjust the command_say() function
accordingly.
Jeremy Stanley [Thu, 21 Apr 2016 14:35:13 +0000 (14:35 +0000)]
Implement new style element support
Add support for new style elements implemented as a flat associative
array of facets prefixed with their element names, and switching to
"." as a prefix separator. This is initially backward compatible
using a toggle and some sniffing of key names to distinguish old and
new style elements. Once the transition from old to new style has
been completed, the compatibility code can be cleaned up.
Jeremy Stanley [Sat, 16 Apr 2016 03:10:32 +0000 (03:10 +0000)]
Test the show element command
Add a functional test for the show element command, to prevent
future regressions in it.
Jeremy Stanley [Sat, 16 Apr 2016 02:06:12 +0000 (02:06 +0000)]
Simplify multiline matching in functional tests
To ease matching across more complex outputs in functional test
conversations, use compiled expressions with the DOTALL flag so that
the "." wildcard will also match newline characters in the stream.
Jeremy Stanley [Sat, 16 Apr 2016 01:02:48 +0000 (01:02 +0000)]
Fix show element command for Py3K
Under Python 3, the dict.keys() method returns a generator rather
than a list and so cannot support the list.sort() method. Instead
use the sorted() function for this purpose in the show element
command. Also switch from concatenation to formatting to embed
facets in the output since they may contain nontext values.
Jeremy Stanley [Thu, 17 Mar 2016 10:03:45 +0000 (10:03 +0000)]
Make escape_macros no-op on nontext values
Update the escape_macros() function to deal with YAML style
Element.get() that may return datatypes other than str. This allows
for indiscriminately passing facets through the function without
recasting them.
Jeremy Stanley [Mon, 7 Mar 2016 13:16:38 +0000 (13:16 +0000)]
Check that functional tests log no errors
After all other functional tests, check the output of show log 7 to
make sure that no earlier tests caused errors at or above that
loglevel.
Jeremy Stanley [Mon, 7 Mar 2016 12:26:43 +0000 (12:26 +0000)]
Test that the show log command works
Exercise the show log command in functional testing, and confirm
that it returns at least some log data.
Jeremy Stanley [Mon, 7 Mar 2016 11:43:13 +0000 (11:43 +0000)]
Fix show log command for Py3K
The behavior of lambda filters has changed in Python 3 such that you
can no longer treat them directly as lists. Trivially reimplement
them as list comprehensions. Also correct a missing minimum loglevel
fallback default.
Jeremy Stanley [Tue, 23 Feb 2016 22:32:27 +0000 (22:32 +0000)]
Fix recursive BrokenPipeException on disconnect
When an admin user has a broken pipe, attempting to log
BrokenPipeException errors to their socket raises another until the
max recursion depth is hit. Set the current user's state to
disconnecting before logging the error instead of after, so as to
avoid this loop.
Jeremy Stanley [Sun, 21 Feb 2016 21:45:07 +0000 (21:45 +0000)]
Switch basic data loader unit from element to node
In preparation for a migration to a new flat data model where
elements are inferred from facet names, stop referring to the
top-level keys in data files as elements and instead call them
nodes. During the transitional period, a node could be an old-style
element or a new-style element's facet.
Jeremy Stanley [Sun, 24 Jan 2016 19:59:45 +0000 (19:59 +0000)]
Explicitly invoke Python 3.x in shebang lines
Since the codebase relies on Py3K-only syntax in places, explicitly
invoke the python3 interpreter rather than expecting the
environment's primary interpreter to support Python 3.x code.
Jeremy Stanley [Mon, 28 Dec 2015 13:13:09 +0000 (13:13 +0000)]
Test admin account restrictions
Add an admin session in the functional tests, and confirm that an
admin can see a restricted function in the help output while a
non-admin cannot see it. Further confirm that an admin command
behaves as an unknown command when a non-admin attempts to call it.
Jeremy Stanley [Tue, 1 Dec 2015 01:45:58 +0000 (01:45 +0000)]
More thorough match on chat mode test
Use a multi-line match on the chat mode functional test, to confirm
both the command feedback and prompt change are present.
Jeremy Stanley [Thu, 19 Nov 2015 04:21:40 +0000 (04:21 +0000)]
Test actor movement
Exercise the move command in all configured directions in the
functional tests.
Jeremy Stanley [Sun, 11 Oct 2015 04:29:32 +0000 (04:29 +0000)]
Mention test session capture log file names
When writing functional test session capture logs, mention the file
names for ease of reference.
Jeremy Stanley [Wed, 7 Oct 2015 20:45:05 +0000 (20:45 +0000)]
Add timing data to test output
Print individual and aggregate functional test durations.
Jeremy Stanley [Sat, 19 Sep 2015 05:09:22 +0000 (05:09 +0000)]
Test chat mode
Add a functional test for chat mode.
Jeremy Stanley [Mon, 7 Sep 2015 03:13:31 +0000 (03:13 +0000)]
Add troubleshooting note to test failure output
When a functional test failure is encountered, recommend the
corresponding log to check for the actual output when an expected
match is not found.
Jeremy Stanley [Sun, 6 Sep 2015 18:18:58 +0000 (18:18 +0000)]
Break outer loop of functional tests on failure
When a test failure is encountered in the inner loop of functional
tests, also break out of the outer loop so that subsequent tests are
not run.
Jeremy Stanley [Tue, 1 Sep 2015 03:54:17 +0000 (03:54 +0000)]
Be more explicit with periods in test matches
Correct some missed escapes for periods (.) in functional test match
strings.
Jeremy Stanley [Sun, 23 Aug 2015 16:20:40 +0000 (16:20 +0000)]
Modularize functional testing
Break functional test dialogue groups into separate modules and add
a description to each for ease of debugging.
Jeremy Stanley [Sat, 22 Aug 2015 19:36:16 +0000 (19:36 +0000)]
Test sentence capitalization
Add a functional test to make sure uncapitalized actor speech gets
displayed with proper sentence capitalization.
Jeremy Stanley [Sun, 9 Aug 2015 19:07:32 +0000 (19:07 +0000)]
Test typo replacements
In the functional test dialogue, add tests for typo replacements.
Jeremy Stanley [Fri, 7 Aug 2015 05:51:03 +0000 (05:51 +0000)]
Test implicit punctuation
In the functional test dialogue, confirm that the configured default
punctuation is appended when implicit punctuation is expected.
Jeremy Stanley [Wed, 15 Jul 2015 20:54:53 +0000 (20:54 +0000)]
Read pending bytes before proceeding with test
After a match in found in each step of the functional test, read in
the remaining buffer before replying. This makes the captures
slightly more readable, especially if the match in early in a block
of output.
Jeremy Stanley [Tue, 2 Jun 2015 06:08:02 +0000 (06:08 +0000)]
Fix punctuation match order and add tests
Add explicit punctuation functional tests with a conversation
between the two test sessions, and fix a nondeterministic bug this
exposed. Punctuation is now matched in a sorted order rather than
arbitrary hash table lookup order, and the sort is reversed so that
the longest match always wins.
Jeremy Stanley [Sun, 17 May 2015 13:24:34 +0000 (13:24 +0000)]
Test interaction between multiple connections
Extend functional testing to utilize a second connection, and test
that the second actor entering and exiting the World causes the
first actor to see the correct messages for each of these events.
Jeremy Stanley [Sat, 9 May 2015 11:45:03 +0000 (11:45 +0000)]
Lay groundwork for more thorough functional tests
Make bin/test support an arbitrary number of connections, be more
verbose about what it's doing, and archive complete captures of each
session involved.
Jeremy Stanley [Sat, 18 Apr 2015 06:54:07 +0000 (06:54 +0000)]
Switch to yaml.safe_load for better security
Use the yaml.safe_load to avoid unwanted privilege escalation due to
deserializing unsafe objects. Also switch to yaml.safe.dump for
symmetry, so that we don't write out files we'll later refuse to
parse.
Jeremy Stanley [Mon, 23 Mar 2015 07:05:55 +0000 (07:05 +0000)]
Be more strict about file permission masks
Set a reasonably strict umask of 0022 when creating most files and
directories. Also set the umask to 0077 for private files to avoid a
brief race where someone could open them for reading after creation
but prior to the prophylactic chmod to 0600.
Jeremy Stanley [Mon, 23 Mar 2015 06:46:05 +0000 (06:46 +0000)]
Accumulate logs if necessary while loading data
Prior to existence of the global universe, the destination files for
logging may not be known. Accumulate log entries during
DataFile.load() if immediate logging fails, so that they can be
flushed later once logging is possible.
Jeremy Stanley [Wed, 18 Mar 2015 09:52:01 +0000 (09:52 +0000)]
Remove stray non-generic references in sample data
The sample data was adapted from an actual test implementation, and
a few references to that original dataset missed cleanup. Take care
of them.
Jeremy Stanley [Wed, 18 Feb 2015 03:35:59 +0000 (03:35 +0000)]
Switch password handler to passlib's PBKDF2
The passlib implementation of PBKDF2 is strong, portable and more
heavily audited. Use that instead of implementing our own custom
handler. Also simplify password use by dropping optional parameters
from the create and verify functions, and don't bother carrying the
old upgrade_legacy_hash public function forward; it can be reworked
to provide in-place hash upgrades later if desired.
Jeremy Stanley [Sat, 17 Jan 2015 01:16:13 +0000 (01:16 +0000)]
Write files directly from yaml.dump
Simplify data file writing by passing a file descriptor directly to
the yaml.dump method rather than unnecessarily passing a string back
out of it and into a separate writer.
Jeremy Stanley [Fri, 9 Jan 2015 21:19:05 +0000 (21:19 +0000)]
Remove UTF-8 encoding header from Python files
The Python 3.x interpreter assumes source files are in UTF-8 by
default, so file encoding headers are no longer necessary. Get rid
of them.
Jeremy Stanley [Mon, 29 Dec 2014 20:21:49 +0000 (20:21 +0000)]
Fix type mismatches in save file rotation
The numeric suffix on rotated save files needs to be treated as an
int during rotation calculations but as a str during filename
generation. Make it so.
Jeremy Stanley [Fri, 26 Dec 2014 10:00:06 +0000 (10:00 +0000)]
Correct avatar inheritance
The inherit facet should be a list, so correct an instance where it
was inadvertently left as a string literal instead.
Jeremy Stanley [Thu, 6 Nov 2014 01:41:56 +0000 (01:41 +0000)]
Log missing avatars
If an avatar is missing, it's likely that it was created but not yet
saved during a crash. This is usually safe to skip but could imply
data corruption, so log the situation with a high log level.
Jeremy Stanley [Mon, 13 Oct 2014 03:12:08 +0000 (03:12 +0000)]
Accumulate logs during setup
Prior to existence of the global universe, the destination files for
logging may not be known. Accumulate log entries during
Universe.load() and then flush them once it's safe to do so.
Jeremy Stanley [Thu, 18 Sep 2014 23:31:35 +0000 (23:31 +0000)]
Add a .gitignore file
Use .gitignore to track ephemeral files created during testing.
Jeremy Stanley [Sat, 13 Sep 2014 06:48:51 +0000 (06:48 +0000)]
Finalize INI to YAML conversion
Convert the main configuration file to YAML and rip out any remnants
of INI support.
Jeremy Stanley [Wed, 6 Aug 2014 14:38:51 +0000 (14:38 +0000)]
Remove the git2gch utility
The git2gch utility was previously used to generate GNU-style
ChangeLog files from Git commit history, but there are useful
existing tools for this now and it's out of scope for mudpy.
Jeremy Stanley [Tue, 5 Aug 2014 00:33:41 +0000 (00:33 +0000)]
Clean up parenthetical spacing in bin/mudpy
For better PEP-8 conformance, correct parenthetical spacing in the
bin/mudpy file.
Jeremy Stanley [Tue, 29 Jul 2014 19:37:50 +0000 (19:37 +0000)]
Mark unused import false positives
Since the reload framework in the package's init imports modules
which look as if they don't get used, add noqa comments inline to
indicate they should be ignored by static analysis tools.
Jeremy Stanley [Mon, 23 Jun 2014 22:35:01 +0000 (22:35 +0000)]
Convert sample data directory index to YAML
Jeremy Stanley [Tue, 17 Jun 2014 05:36:27 +0000 (05:36 +0000)]
Support data directory indices in YAML
* lib/mudpy/data.py(DataFile.load_yaml,DataFile.load_mpy): Add routines
to look for a YAML file index as well as the old INI-based format. For
YAML files, don't bother recasting entries to lists or dicts since they
should already be returned as the desired type.