This switches the logger to Debug. By default the library logs things
internally despite returning errors. Due to the way we had it hooked up
this resulted in scary error level logs being emitted for innocuous
errors. Now those errors will only be visible in debug mode.
Fixes: #1398
Relates to: #1336
When GTS is running in a container runtime which has configured CPU or
memory limits or under an init system that uses cgroups to impose CPU
and memory limits the values the Go runtime sees for GOMAXPROCS and
GOMEMLIMIT are still based on the host resources, not the cgroup.
At least for the throttling middlewares which use GOMAXPROCS to
configure their queue size, this can result in GTS running with values
too big compared to the resources that will actuall be available to it.
This introduces 2 dependencies which can pick up resource contraints
from the current cgroup and tune the Go runtime accordingly. This should
result in the different queues being appropriately sized and in general
more predictable performance. These dependencies are a no-op on
non-Linux systems or if running in a cgroup that doesn't set a limit on
CPU or memory.
The automatic tuning of GOMEMLIMIT can be disabled by either explicitly
setting GOMEMLIMIT yourself or by setting AUTOMEMLIMIT=off. The
automatic tuning of GOMAXPROCS can similarly be counteracted by setting
GOMAXPROCS yourself.
* [feature] Add throttling middleware to AP endpoints
* refactor a lil bit
* use config setting, start updating docs
* doc updates
* use relative links in faq doc
* small docs fixes
* return code 503 instead of 429 when throttled
* throttle other endpoints too
* simplify token channel prefills
* interim commit: start refactoring middlewares into package under router
* another interim commit, this is becoming a big job
* another fucking massive interim commit
* refactor bookmarks to new style
* ambassador, wiz zeze commits you are spoiling uz
* she compiles, we're getting there
* we're just normal men; we're just innocent men
* apiutil
* whoopsie
* i'm glad noone reads commit msgs haha :blob_sweat:
* use that weirdo go-bytesize library for maxMultipartMemory
* fix media module paths
* move caches to a separate State{} structure
Signed-off-by: kim <grufwub@gmail.com>
* fix call to log.Panic not using formatted call
Signed-off-by: kim <grufwub@gmail.com>
* move caches to use interfaces, to make switchouts easier in future
Signed-off-by: kim <grufwub@gmail.com>
* fix rebase issue
Signed-off-by: kim <grufwub@gmail.com>
* improve code comment
Signed-off-by: kim <grufwub@gmail.com>
* fix further issues after rebase
Signed-off-by: kim <grufwub@gmail.com>
* heh
Signed-off-by: kim <grufwub@gmail.com>
* add missing license text
Signed-off-by: kim <grufwub@gmail.com>
Signed-off-by: kim <grufwub@gmail.com>
* [feature] overhaul the oidc system
this allows for more flexible username handling and prevents account
takeover using old email addresses
* [feature] add migration path for old OIDC users
* [feature] nicer error reporting for users
* [docs] document the new OIDC flow
* [fix] return early on oidc error
* [docs]: add comments on the finalization logic
Implements #864 and should speed up s3 based installations by a lot.
With more static urls, we can then also implement #1026 for even
better performance when used in conjunction with CDNs
* add FilePath regex
* add `admin media prune orphaned` command
* add prune orphaned function to media manager
* don't mark flag as required
* document admin media prune orphaned cmd
* oh envparsing.sh you coy minx
The NewSignup method was already being called with
requireApproval=false, but it had emailVerified=false as well, which
meant that it was required to use the `admin account confirm` command
to verify the email before the newly-created user could log in.
I think that was probably an oversight; effectively it did require
approval anyway. Changing emailVerified to true allows you to just
create the account and log in immediately, reducing the opportunity
for manual error to sneak in.
Also updated the docs to remove the mention of needing to confirm new
accounts. However, I've left the confirmation command alone because I
think once we have web signups, it will be needed in that context.
* convert most of the caches to use result.Cache{}
* add caching of emojis
* fix issues causing failing tests
* update go-cache/v2 instances with v3
* fix getnotification
* add a note about the left-in StatusCreate comment
* update EmojiCategory db access to use new result.Cache{}
* fix possible panic in getstatusparents
* further proof that kim is not stinky
* bump go-store version to v2.0.5, init kv.KVStore without initial clean (as we are using for storage, not as a key-value store)
Signed-off-by: kim <grufwub@gmail.com>
* remove newline
Signed-off-by: kim <grufwub@gmail.com>
Signed-off-by: kim <grufwub@gmail.com>
* use bun.Ident for user queries
* use bun.Ident for account queries
* use bun.Ident for media queries
* add DeleteAccount func
* remove CaseInsensitive in Where+use Ident ipv Safe
* update admin db
* update domain, use ident
* update emoji, use ident
* update instance queries, use bun.Ident
* fix media
* update mentions, use bun ident
* update relationship + tests
* use tableexpr
* add test follows to bun db test suite
* update notifications
* updatebyprimarykey => updatebyid
* fix session
* prefer explicit ID to pk
* fix little fucky wucky
* remove workaround
* use proper db func for attachment selection
* update status db
* add m2m entries in test rig
* fix up timeline
* go fmt
* fix status put issue
* update GetAccountStatuses
* update dependencies, bump Go version to 1.19
* bump test image Go version
* update golangci-lint
* update gotosocial-drone-build
* sign
* linting, go fmt
* update swagger docs
* update swagger docs
* whitespace
* update contributing.md
* fuckin whoopsie doopsie
* linterino, linteroni
* fix followrequest test not starting processor
* fix other api/client tests not starting processor
* fix remaining tests where processor not started
* bump go-runners version
* don't check last-webfingered-at, processor may have updated this
* update swagger command
* update bun to latest version
* fix embed to work the same as before with new bun
Signed-off-by: kim <grufwub@gmail.com>
Co-authored-by: tsmethurst <tobi.smethurst@protonmail.com>
* validate web-asset-base-dir
* move default icons into converter
* always ensure avatar + header on api accounts
* update tests
* add default header
* don't return error from web module creation anymore
* tidy a bit
* use pngs for default avatars rather than svgs
* upstep bun and sqlite versions
* allow specific columns to be updated in the db
* only update necessary columns for user
* bit tidier
* only update necessary fields of media_attachment
* only update relevant instance fields
* update tests
* update only specific account columns
* use bool pointers on gtsmodels
includes attachment, status, account, user
* update columns more selectively
* test all default fields on new account insert
* updating remaining bools on gtsmodels
* initialize pointer fields when extracting AP emoji
* copy bools properly
* add copyBoolPtr convenience function + test it
* initialize false bool ptrs a bit more neatly
* first commit
Signed-off-by: kim <grufwub@gmail.com>
* replace logging with our own log library
Signed-off-by: kim <grufwub@gmail.com>
* fix imports
Signed-off-by: kim <grufwub@gmail.com>
* fix log imports
Signed-off-by: kim <grufwub@gmail.com>
* add license text
Signed-off-by: kim <grufwub@gmail.com>
* fix package import cycle between config and log package
Signed-off-by: kim <grufwub@gmail.com>
* fix empty kv.Fields{} being passed to WithFields()
Signed-off-by: kim <grufwub@gmail.com>
* fix uses of log.WithFields() with whitespace issues and empty slices
Signed-off-by: kim <grufwub@gmail.com>
* *linter related grumbling*
Signed-off-by: kim <grufwub@gmail.com>
* gofmt the codebase! also fix more log.WithFields() formatting issues
Signed-off-by: kim <grufwub@gmail.com>
* update testrig code to match new changes
Signed-off-by: kim <grufwub@gmail.com>
* fix error wrapping in non fmt.Errorf function
Signed-off-by: kim <grufwub@gmail.com>
* add benchmarking of log.Caller() vs non-cached
Signed-off-by: kim <grufwub@gmail.com>
* fix syslog tests, add standard build tags to test runner to ensure consistency
Signed-off-by: kim <grufwub@gmail.com>
* make syslog tests more robust
Signed-off-by: kim <grufwub@gmail.com>
* fix caller depth arithmatic (is that how you spell it?)
Signed-off-by: kim <grufwub@gmail.com>
* update to use unkeyed fields in kv.Field{} instances
Signed-off-by: kim <grufwub@gmail.com>
* update go-kv library
Signed-off-by: kim <grufwub@gmail.com>
* update libraries list
Signed-off-by: kim <grufwub@gmail.com>
* fuck you linter get nerfed
Signed-off-by: kim <grufwub@gmail.com>
Co-authored-by: tobi <31960611+tsmethurst@users.noreply.github.com>
* feat: vendor minio client
* feat: introduce storage package with s3 support
* feat: serve s3 files directly
this saves a lot of bandwith as the files are fetched from the object
store directly
* fix: use explicit local storage in tests
* feat: integrate s3 storage with the main server
* fix: add s3 config to cli tests
* docs: explicitly set values in example config
also adds license header to the storage package
* fix: use better http status code on s3 redirect
HTTP 302 Found is the best fit, as it signifies that the resource
requested was found but not under its presumed URL
307/TemporaryRedirect would mean that this resource is usually located
here, not in this case
303/SeeOther indicates that the redirection does not link to the
requested resource but to another page
* refactor: use context in storage driver interface
* move finger to dereferencer
* totally break GetRemoteAccount
* start reworking finger func a bit
* start reworking getRemoteAccount a bit
* move mention parts to namestring
* rework webfingerget
* use util function to extract webfinger parts
* use accountDomain
* rework finger again, final form
* just a real nasty commit, the worst
* remove refresh from account
* use new ASRepToAccount signature
* fix incorrect debug call
* fix for new getRemoteAccount
* rework GetRemoteAccount
* start updating tests to remove repetition
* break a lot of tests
Move shared test logic into the testrig,
rather than having it scattered all over
the place. This allows us to just mock
the transport controller once, and have
all tests use it (unless they need not to
for some other reason).
* fix up tests to use main mock httpclient
* webfinger only if necessary
* cheeky linting with the lads
* update mentionName regex
recognize instance accounts
* don't finger instance accounts
* test webfinger part extraction
* increase default worker count to 4 per cpu
* don't repeat regex parsing
* final search for discovered accountDomain
* be more permissive in namestring lookup
* add more extraction tests
* simplify GetParseMentionFunc
* skip long search if local account
* fix broken test
* start fiddling about with oauth server
* start returning more helpful errors from oauth
* test helpful(ish) token errors
* add missing license header
* update templates
* start reworking api error handling
* update template
* return AP status at web endpoint if negotiated
* start making api error handling much more consistent
* update account endpoints to new error handling
* use new api error handling in admin endpoints
* go fmt ./...
* use api error logic in app
* use generic error handling in auth
* don't export generic error handler
* don't defer clearing session
* user nicer error handling on oidc callback handler
* tidy up the sign in handler
* tidy up the token handler
* use nicer error handling in blocksget
* auth emojis endpoint
* fix up remaining api endpoints
* fix whoopsie during login flow
* regenerate swagger docs
* change http error logging to debug
* move config flag names and usage to config package, rewrite config package to use global Configuration{} struct
Signed-off-by: kim <grufwub@gmail.com>
* improved code comment
Signed-off-by: kim <grufwub@gmail.com>
* linter
Signed-off-by: kim <grufwub@gmail.com>
* fix unmarshaling
Signed-off-by: kim <grufwub@gmail.com>
* remove kim's custom go compiler changes
Signed-off-by: kim <grufwub@gmail.com>
* generate setter and flag-name functions, implement these in codebase
Signed-off-by: kim <grufwub@gmail.com>
* update deps
Signed-off-by: kim <grufwub@gmail.com>
* small change
Signed-off-by: kim <grufwub@gmail.com>
* appease the linter...
Signed-off-by: kim <grufwub@gmail.com>
* move configuration into ConfigState structure, ensure reloading to/from viper settings to keep in sync
Signed-off-by: kim <grufwub@gmail.com>
* lint
Signed-off-by: kim <grufwub@gmail.com>
* update code comments
Signed-off-by: kim <grufwub@gmail.com>
* fix merge issue
Signed-off-by: kim <grufwub@gmail.com>
* fix merge issue
Signed-off-by: kim <grufwub@gmail.com>
* improved version string (removes time + go version)
Signed-off-by: kim <grufwub@gmail.com>
* fix version string build to pass test script + consolidate logic in func
Signed-off-by: kim <grufwub@gmail.com>
* add license text, update config.Defaults comment
Signed-off-by: kim <grufwub@gmail.com>
* add license text to generated config helpers file
Signed-off-by: kim <grufwub@gmail.com>
* defer unlock on config.Set___(), to ensure unlocked on panic
Signed-off-by: kim <grufwub@gmail.com>
* make it more obvious which cmd flags are being attached
Signed-off-by: kim <grufwub@gmail.com>
* add optional config validation
* clarify that host and protocol are required
* add validation for host and protocol
* pass prerunArgs as a struct (validate by default)
* rename + tidy up remote pruning
* fix media attachment account join
see https://bun.uptrace.dev/guide/golang-orm.html#table-relationships
* update logging to new function name
* add get avatars and headers to bun
* add pruneallmeta function
* don't set uncached since we're deleting anyway
* fix totalPruned being in wrong place
* test pruning meta
* go fmt ./...
* rename mediaprune
* add meta pruning to routine mediaprune
* tidy up cleanup job scheduling
* rename adminmediaremoteprune
* update mediacleanup to use renamed prune func
* update swagger docs a little bit
* reuse cancel + context
* cache transports in controller by privkey-generated pubkey, add retry logic to transport requests
Signed-off-by: kim <grufwub@gmail.com>
* update code comments, defer mutex unlocks
Signed-off-by: kim <grufwub@gmail.com>
* add count to 'performing request' log message
Signed-off-by: kim <grufwub@gmail.com>
* reduce repeated conversions of same url.URL object
Signed-off-by: kim <grufwub@gmail.com>
* move worker.Worker to concurrency subpackage, add WorkQueue type, limit transport http client use by WorkQueue
Signed-off-by: kim <grufwub@gmail.com>
* fix security advisories regarding max outgoing conns, max rsp body size
- implemented by a new httpclient.Client{} that wraps an underlying
client with a queue to limit connections, and limit reader wrapping
a response body with a configured maximum size
- update pub.HttpClient args passed around to be this new httpclient.Client{}
Signed-off-by: kim <grufwub@gmail.com>
* add httpclient tests, move ip validation to separate package + change mechanism
Signed-off-by: kim <grufwub@gmail.com>
* fix merge conflicts
Signed-off-by: kim <grufwub@gmail.com>
* use singular mutex in transport rather than separate signer mus
Signed-off-by: kim <grufwub@gmail.com>
* improved useragent string
Signed-off-by: kim <grufwub@gmail.com>
* add note regarding missing test
Signed-off-by: kim <grufwub@gmail.com>
* remove useragent field from transport (instead store in controller)
Signed-off-by: kim <grufwub@gmail.com>
* shutup linter
Signed-off-by: kim <grufwub@gmail.com>
* reset other signing headers on each loop iteration
Signed-off-by: kim <grufwub@gmail.com>
* respect request ctx during retry-backoff sleep period
Signed-off-by: kim <grufwub@gmail.com>
* use external pkg with docs explaining performance "hack"
Signed-off-by: kim <grufwub@gmail.com>
* use http package constants instead of string method literals
Signed-off-by: kim <grufwub@gmail.com>
* add license file headers
Signed-off-by: kim <grufwub@gmail.com>
* update code comment to match new func names
Signed-off-by: kim <grufwub@gmail.com>
* updates to user-agent string
Signed-off-by: kim <grufwub@gmail.com>
* update signed testrig models to fit with new transport logic (instead uses separate signer now)
Signed-off-by: kim <grufwub@gmail.com>
* fuck you linter
Signed-off-by: kim <grufwub@gmail.com>
* linting with new golangci-lint version
* update go to 1.18
* bump versions in drone.yml
* use new runtime/debug package for version info
* remove Commit build flag from goreleaser
* remove mock commit + version from build script
* go fmt
* add dummy version env flag to test container
* install git in golang container for testing
* only set versionString if Version is defined
* Add whereNotEmptyAndNotNull
* Add GetRemoteOlderThanDays
* Add GetRemoteOlderThanDays
* Add PruneRemote to Manager interface
* Start implementing PruneRemote
* add new attachment + status to tests
* fix up and test GetRemoteOlderThan
* fix bad import
* PruneRemote: return number pruned
* add Cached column to mediaattachment
* update + test pruneRemote
* update mediaTest
* use Cached column
* upstep bun to latest version
* embed structs in mediaAttachment
* migrate mediaAttachment to new format
* don't default cached to true
* select only remote media
* update db dependencies
* step bun back to last working version
* update pruneRemote to use Cached field
* fix storage path of test attachments
* add recache logic to manager
* fix trimmed aspect ratio
* test prune and recache
* return errwithcode
* tidy up different paths for emoji vs attachment
* fix incorrect thumbnail type being stored
* expose TransportController to media processor
* implement tee-ing recached content
* add thoughts of dog to test fedi attachments
* test get remote files
* add comment on PruneRemote
* add postData cleanup to recache
* test thumbnail fetching
* add incredible diagram
* go mod tidy
* buffer pipes for recache streaming
* test for client stops reading after 1kb
* add media-remote-cache-days to config
* add cron package
* wrap logrus so it's available to cron
* start and stop cron jobs gracefully
* add preparable and timelineable interfaces
* initialize timeline manager within the processor
* generic renaming
* move status-specific timeline logic into the processor
* refactor timeline to make it useful for more than statuses
