* Use sessionmaker instead of context block
Essentially there were two bugs:
1) The logging lines before explicit `Session.begin` caused the root
"A transaction is already begun on this Session.", which is
resolved by moving the logging within the explicit session block,
as intended.
2) We were reusing a Session across cron runs. The architecture was
intended to not do so, but we didn't have an actual sessionmaker
available to cron. We probably ought to get rid of scoped_session
elsewhere, but baby steps to resolve the immediate issue.
In testing, this works correctly with the fix to `db.begin`, and
tests with deliberately reintroducing the logging bug, the changes
to session creation prevent the every-15sec stack trace spam.
---------
Almost-entirely-authored-by: TLSM <duolsm@outlook.com>
b46ada9f72 added support for dotenv files to be loaded in lieu of
actual environment variables. This code has since been neglected across
infrastructure changes that removed the need for an env file, moved
its location, and changed dependency management style.
This won't be operative in production (and already isn't, since there'd
be an ImportError on 'dotenv' if so) but makes some local testing
setups more convenient.
* prepare codebase to create scheduled tasks
there is some prep work involved with this. the scheduler would be happy
if this work was done. simply, we extract out the `created_utc`
interface from *everything* that uses it such that we don't have to
repeat ourselves a bunch. all fun stuff.
next commit is the meat of it.
* cron: basic backend work for scheduler
* avoid ipmort loop
* attempt 2 at fixing import loops
* parathensize because operator precedence
* delete file that came back for some reason.
* does NOPing the oauth apps work?
* import late and undo clients.py change
* stringify column names.
* reorder imports.
* remove task reference
* fix missing mapper object
* make coupled to repeatabletask i guess
* sanitize: fix sanitize imports
* import shadowing crap
* re-shadow shadowed variable
* fix regexes
* use the correct not operator
* readd missing commit
* scheduler: SQLA only allows concrete relations
* implement submission scheduler
* fix import loop with db_session
* get rid of import loop in submission.py and comment.py
* remove import loops by deferring import until function clal
* i give up.
* awful.
* ...
* fix another app import loop
* fix missing import in route handler
* fix import error in wrappers.py
* fix wrapper error
* call update wrapper in the admin_level_required case
* :marseyshrug:
* fix issue with wrapper
* some cleanup and some fixes
* some more cleanup
let's avoid polluting scopes where we can.
* ...
* add SCHEDULED_POSTS permission.
* move const.py into config like the other files.
* style fixes.
* lock table for concurrency improvements
* don't attempt to commit on errors
* Refactor code, create `TaskRunContext`, create python callable task type.
* use import contextlib
* testing stuff i guess.
* handle repeatable tasks properly.
* Attempt another fix at fighting the mapper
* do it right ig
* SQLA1.4 doesn't support nested polymorphism ig
* fix errenous class import
* fix mapper errors
* import app in wrappers.py
* fix import failures and stuff like that.
* embed and import fixes
* minor formatting changes.
* Add running state enum and don't attempt to check for currently running tasks.
* isort
* documentation, style, and commit after each task.
* Add completion time and more docs, rename, etc
* document `CRON_SLEEP_SECONDS` better.
* add note about making LiteralString
* filter out tasks that have been run in the future
* reference RepeatableTask's `__tablename__` directly
* use a master/slave configuration for tasks
the master periodically checks to see if the slave is alive, healthy,
and not taking too many resources, and if applicable kills its
child and restarts it.
only one relation is supported at the moment.
* don't duplicate process unnecessarily
* note impl detail, add comments
* fix imports.
* getting imports to stop being stupid.
* environment notes.
* syntax derp
* *sigh*
* stupid environment stuff
* add UI for submitting a scheduled post
* stupid things i need to fix the user class
* ...
* fix template
* add formkey
* pass v
* add hour and minute field
* bleh
* remove concrete
* the sqlalchemy docs are wrong
* fix me being dumb and not understanding error messages
* missing author attribute for display
* author_name property
* it's a property
* with_polymorphic i think fixes this
* dsfavgnhmjk
* *sigh*
* okay try this again
* try getting rid of the comment section
* include -> extends
* put the div outside of the thing.
* fix user page listings :/
* mhm
* i hate this why isn't this working
* this should fix it
* Fix posts being set as disabled by default
* form UI imrpovements
* label
* <textarea>s should have their closing tag
* UI fixes.
* and fix errenous spinner thing.
* don't abort(415) when browsers send 0 length files for some reason
* UI improvements
* line break.
* CSS :S
* better explainer
* don't show moderation buttons for scheduled posts
* ...
* meh
* add edit form
* include forms on default page.
* fix hour minute selectino.
* improve ui i guess and add api
* Show previous postings on scheduled task page
* create task id
* sqla
* posts -> submissions
* fix OTM relationship
* edit URL
* use common formkey control
* Idk why this isn't working
* Revert "Idk why this isn't working"
This reverts commit 3b93f741df.
* does removing viewonly fix it?
* don't import routes on db migrations
* apparently this has to be a string
* UI improvements redux
* margins and stuff
* add cron to supervisord
* remove stupid duplication
* typo fix
* postgres syntax error
* better lock and error handling
* add relationship between task and runs
* fix some ui stuff
* fix incorrect timestamp comparison
* ...
* Fix logic errors blocking scheduled posts
Two bugs here:
- RepeatableTask.run_time_last <= now: run_time_last is NULL by
default. NULL is not greater than, less than, or equal to any
value. We use NULL to signify a never-run task; check for that
condition when building the task list.
- `6 <= weekday <= 0`: there is no integer that is both gte 6 and
lte 0. This was always false.
* pasthrough worker process STDOUT and STDERR
* Add scheduler to admin panel
* scheduler
* fix listing and admin home
* date formatting ixes
* fix ages
* task user interface
* fix some more import crap i have to deal with
* fix typing
* avoid import loop
* UI fixes
* fix incorrect type
* task type
* Scheduled task UI improvements (add runs and stuff)
* make the width a lil bit smaller
* task runs.
* fix submit page
* add alembic migration
* log on startup
* Fix showing edit button
* Fix logic for `can_edit` (accidentally did `author_id` instead of `id`)
* Broad review pass
Review:
- Call `invalidate_cache` with `is_html=` explicitly for clarity,
rather than a bare boolean in the call args.
- Remove `marseys_const*` and associated stateful const system:
the implementation was good if we needed them, but TheMotte
doesn't use emoji, and a greenfield emoji system would likely
not keep those darned lists floating in thread-local scope.
Also they were only needed for goldens and random emoji, which
are fairly non-central features.
- Get `os.environ` fully out of the templates by using the new
constants we already have in files.helpers.config.environment.
- Given files.routes.posts cleanup,get rid of shop discount dict.
It's already a mapping of badge IDs to discounts for badges that
likely won't continue to exist (if they even do at present).
- RepeatableTaskRun.exception: use `@property.setter` instead of
overriding `__setattr__`.
Fix:
- Welcome message literal contained an indented Markdown code block.
- Condition to show "View source" button changed to show source to
logged out. This may well be a desirable change, but it's not
clearly intended here.
* Fix couple of routing issues
* fix 400 with post body editing
* Add error handler for HTTP 415
* fix router giving wrong arg name to handler
* Use supervisord to monitor memory rather than DIY
Also means we're using pip for getting supervisord now, so we don't rely
on the Debian image base for any packages.
* fix task run elapsed time display
* formatting and removing redundant code
* Fix missing ModAction import
* dates and times fixes
* Having to modify imports here anyway, might as
well change it.
* correct documentation.
* don't use urlunparse
* validators: import sanitize instead of from syntax
* cron: prevent races on task running
RepeatableTask.run_state_enum acts as the mutex on repeatable tasks.
Previously, the list of tasks to run was acquired before individually
locking each task. However, there was a period where the table is both
unlocked and the tasks are in state WAITING between those points.
This could potentially have led to two 'cron' processes each running the
same task simultaneously. Instead, we check for runnability both when
building the preliminary list and when mutexing the task via run state
in the database.
Also:
- g.db and the cron db object are both instances of `Session`, not
`scoped_session` because they are obtained from
`scoped_session.__call__`, which acts as a `Session` factory.
Propagate this to the type hints.
- Sort order of task run submissions so /tasks/scheduled_posts/<id>
"Previous Task Runs" listings are useful.
* Notify followers on post publication
This was old behavior lost in the refactoring of the submit endpoint.
Also fix an AttributeError in `Follow.__repr__` which carried over
from all the repr copypasta.
* Fix image attachment
Any check for `file.content_length` relies on browsers sending
Content-Length headers with the request. It seems that few actually do.
The pre-refactor approach was to check for truthiness, which excludes
both None and the strange empty strings that we seem to get in absence
of a file upload. We return to doing so.
---------
Co-authored-by: TLSM <duolsm@outlook.com>
We currently use Flask-Limiter 2.7.0. Upgrading to 3.1.0 breaks
because of a breaking change to keyword order. Yes, seriously:
https://flask-limiter.readthedocs.io/en/stable/changelog.html#v3-0-0
Not updating the lockfile here, pending @ZorbaTHut testing some infra
updates; however, this is necessary to be able to do so, and it still
works in 2.7.0.
Ports in lightly modified logic from the upstream which tracks active
sessions to provide counters and listings to understand site traffic
in the admin panel.
Recently, unrelated changes led to enabling logging for flask-limiter
accidentally, at which point it was discovered that it wasn't actually
limiting requests due to Limiter.key_func = get_CF not being proper
for either prod (not behind Cloudflare) or localhost (likewise).
We instead use the remote_addr attached directly to the request using
the existing flask-limiter function to do so.
Detailed troubleshooting at:
https://github.com/themotte/rDrama/issues/222#issuecomment-1229489062
This change disables multimedia embedding:
- In comments and comments replies.
- In new submissions.
- In comment & submission preview
And it's all toggle-able via an envvar, except for the JS bits,
but I linked those to the github issue, so should be easy to find
in the future.
The way it works is:
- removes markdown image/video syntax,
eg. `` into ``
- changes link text into anchors, eg.
`https://example.org/someimage.jpg` into
`[https://example.org/someimage.jpg](https://example.org/someimage.jpg)`
- removes html img/video/audio tags, eg.
`<img href="https://example.org/someimage.jpg" />` into ``
- when embedding gifs via the giphy modal in "new submission", it will
insert only an anchor to the gif
- when attaching an image, it will upload the image, then add only an
anchor to the post/comment body
I tested this manually, but not sure if I got all the test cases. What I
checked was:
- create comment w/ image/video/audio media using markdown -> success
- create comment reply w/ image/video/audio media using markdown ->
success
- create comment w/ link to img/imgur/youtube/audio -> success
- create comment w/ attachment -> success
- create comment reply w/ attachment -> success
- create comment w/ img/video tag -> success
- create comment reply w/ image/video tag -> success
- create post submission w/ image/video/media using markdown -> success
- create post submission w/ link to img/imgur/youtube/audio -> success
- create post submission w/ attachment -> success
- create post submission w/ giphy gif -> success
Also, updated the formatting page.
Co-authored-by: Ben Rog-Wilhelm <zorba-github@pavlovian.net>
This caps the number of mentions in a submission or comment to
MENTION_LIMIT, which is set to 100 by default. When limit is exceeded,
returns a helpful message to the user.
* #39 Add Flask-Migrate dep
* #39 Make it such that flask db init can run
https://github.com/miguelgrinberg/Flask-Migrate/issues/196#issuecomment-381343393
* Run flask db init, update migrations.env, commit artifacts
* Set up a script such that you can `docker-compose exec files bash -c 'cd /service; FLASK_APP="files/cli:app" flask '` and have it do whatever flask thing you want
* Fix circular dependency
* import * is evil
* Initial alembic migration, has issues with constraints and nullable columns
* Bring alts table up to date with alembic autogenerate
* Rerun flask db revision autogenerate
* Bring award_relationships table up to date with alembic autogenerate
* [#39/alembic] files/classes/__init__.py is evil but is at least explicitly evil now
* #39 fix model in files/classes/badges.py
* #39 fix model in files/classes/domains.py and files/classes/clients.py
* #39 fix models: comment saves, comment flags
* #39 fix models: comments
* Few more imports
* #39 columns that are not nullable should be flagged as not nullable
* #39 Add missing indexes to model defs
* [#39] add missing unique constraints to model defs
* [#39] Temporarily undo any model changes which cause the sqlalchemy model to be out of sync with the actual dump
* #39 Deforeignkeyify the correct column to make alembic happy
* #39 flask db revision --autogenerate now creates an empty migration
* #39 Migration format such that files are listed in creation order
* #39 Better first revision
* #39 Revert the model changes that were required to get to zero differences between db revision --autogenerate and the existing schema
* #39 The first real migration
* #39 Ensure that foreign key constraints are named in migration
* #39 Alembic migrations for FK constraints, column defs
* [#39] Run DB migrations before starting tests
* [#39] New test to ensure migrations are up to date
* [#39] More descriptive test failure message
* Add -T flag to docker-compose exec
* [#39] Run alembic migrations when starting the container
I created a new theme, "TheMotte" in /assests/css. Someone with any design
skill at all should try improving this. Currently it is a copy of light.css
__main__.py defines the experience for a user without an account. I removed the
check for an OS environment variable, and defaulted to the new theme.
/classes/user.py defines the default for a new account. I removed the check for an OS
environment variable, and defaulted to the new theme.
/routes/settings.py has a list of the themes that checks them for... something.
I added the new one here, just in case.
/templates/settings_profile.html has a hardcoded list that controls which
themes appear in the user settings page. Added the new theme here, so you can
select it.
