| Commit message (Collapse) | Author | Lines |
|
Some packages (currently only python) should be prioritised to others,
the previous cleanup didn't account for other packages loading before it
which resulted in duplicate symbols getting the python prefix and the
original symbols linking to most probably undesired pages
|
|
We also clear the cache when removing a package
|
|
If we're parsing a page for a symbol that's out of the cache
and encounter a symbol that was already cached we can update that symbol
to keep it up to date without additional requests
|
|
This rework requires us to delete packages caches easily with
deleting the package hash instead of having to pattern match all
keys and delete those.
The interface was also updated to accept DocItems instead of requiring
callers to construct the keys
|
|
|
|
Group name interning was also moved to the DocItem creation
to group the behaviour
|
|
|
|
With redis we need to make sure we don't send the "error"
string into the cache, returning None instead of the string
and then setting it manually in the caller makes this nicer
compared to checking against a string
|
|
|
|
|
|
|
|
|
|
|
|
With the catchall else condition and symbols from FORCE_PREFIX_GROUPS
getting renamed even when being overwritten, we can ignore the package
handling and let it go to the else which adds the package prefix
instead of a group
|
|
Previously weo nly added the package name for symbols
that shared are named name with an another symbol, but
in some edge cases we can get to this point with symbols
that weren't renamed but have name conflicts, causing some
to get overwritten completely without the capturing condition
|
|
|
|
The regex we use to detect codeblocks did not recognize language
specifiers that use a dash, a plus, or a dot in their name. As there are
valid language specifiers, such as python-repl and c++, that use those
characters, I've changed the regex to reflect that.
The character set used now reflects the characters used in language
specifiers in highlight.js.
Signed-off-by: Sebastiaan Zeeff <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Fixes BOT-AN
|
|
|
|
|
|
|
|
Signed-off-by: Daniel Brown <[email protected]>
|
|
Changing a hyphen to an underscore in the config
Co-authored-by: kwzrd <[email protected]>
|
|
Signed-off-by: Daniel Brown <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
It's better to have a self-documenting name than a comment, which, by
the way, was using the old name for the decorator.
|
|
|
|
|
|
|
|
If a user is already connected to a voice channel at the time of getting verified through voice gate, they won't have their permissions actually apply to their current session. This change adds information on verifying so they know they must reconnect to have the changes apply.
|
|
An overlooked grammatical error occurred in exactly 1 (one) of the possible failure reasons when being verified for the voice gate system. This was unacceptable to the masses, so a swift correction has been added to address it, adding 1 (one) additional word to the listed reason.
|
|
When verified users get their role, they cannot see the voice-verification channel anymore, so I've added a 3 second delay for granting the role in order to ensure there's some time for them to see the response.
I've also moved the DM message to only be sent if the DM message succeeds, and to not mention them in-channel to avoid distracting them from the DM notification unnecessarily, as I'm sure they'll see a near-instant response to their command usage in that channel.
|
|
On voiceban, a users effective permissions will change to not allow speaking, however this permission isn't effective until rejoining. To ensure a voiceban is immediately in effect, the user will be disconnected from any voice channels.
|