After some more thought, a response hook should be checking for
whether headers are sent, and executing (or not executing) the
default logic in that case.
Before, we were relying on hooks to call data.next() to continue
execution, but it makes more sense to have the listener either
send a response or not, and handle the behaviour afterwards.
* feat: testing suite integration for openapi spec
The testing suite now takes the openapi spec into account. It will
check each route defined, make a call to it, and compare the
response with the defined schema. Any mismatches will cause the
test to fail.
* fix(openapi): removed debug stuff from tests
* fix(openapi): fixed some tests
* fix(openapi): added additional check to tests, test fixes
* fix(openapi): better tests, fixed spec errors
* fix(openapi): bad conditional in test
* fix: oops
* fix(openapi): more tests fixing
* fix(openapi): more tests
* fix(openapi): fix some more tests
* fix: verbose'd an info log
* fix: topic pagination route returns schema-optimized pagination block
* fix(openapi): more test/spec fixes
* fix(openapi): accidentally sending in authenticated jar for anon routes
* fix(openapi): more test/spec fixes
* fix(openapi): more spec fixes
* fix: timestampReadable Invalid Date
* fix(openapi): more tests... almost there
* fix(openapi): more tests fixing
* fix(openapi): finally all tests passing
* fix(openapi): added reverse test to compare response to spec
... and fixed all the tests that broke
* fix: remove tests related to group covers, as route is gone
* fix(openapi): broken test on travis
* fix(openapi): broken test on travis
* fix(openapi): broken test on travis
* fix(openapi): object cache is not present for psql
* fix: tests
Co-authored-by: Barış Soner Uşaklı <barisusakli@gmail.com>
* feat: support pnpm, cnpm
* feat: store supportedPackageManager in default package.json
* feat: store supportedPackageManager in src/cli/package-install.js
Co-authored-by: a632079 <a632079@qq.com>
* fixes sitemap base url issue when using with subfolder
* fixes sitemap base url issue when using with subfolder
* uses directly relative path
* removes prefix variable
fix#8079
Unified Topology changes the concept of connecting and causes isConnected to always return true, making any reconnection options useless. Next major version of the driver will remove isConnected entirely.
So:
- reconnect tries
- reconnectInterval
- autoReconnect
Are deprecated and don't really do anything anymore (when unified topology is used at least) other than causing warnings at startup since the 3.4 version of the driver (the current version used by NodeBB is 3.5).
Reference: http://mongodb.github.io/node-mongodb-native/3.5/reference/unified-topology/
* fix: #8142, broken site if no server-side session
During the `addHeader` middleware, a check is now done to see if
`req.session.meta` is present. This value is only present if the user
has a valid server-side session. If it is missing, then it is probably
safe to assume that the server-side session was deleted (either
intentionally or accidentally). In that scenario, the client-side cookie
should be cleared.
Also, there was an issue where the sessionRefresh flag was never cleared
after a successful login, so that was fixed too.
* feat: exported method to get cookie config
* fix: don't clear cookie if cookie is being set
* fix: socket.io tests
Co-authored-by: Barış Soner Uşaklı <barisusakli@gmail.com>
In scenarios where onSuccessfulLogin was not called in the SSO plugin,
core's calling of onSuccessfulLogin was prematurely returning, because
it was checking the wrong value.
This commit fixes the issue by checking a different value.
- Now showing user-join message for chat owner in addition to the newly
added user
- If a modal already exists for a room when createModal is called, don't
return null, return the modal
- The logic for the build step now defaults to series instead of
parallel, unless more than 4 CPU cores are detected by the os
library.
- The `--series` flag still exists, and will enforce build in
series, as before.
* feat: add assignee checking when updating flag
Prior to this, it was possible to update the assignee to any value (or
any user. This commit adds checking to allow only admins, global
moderators, or in the case of flagged posts, moderators.
Also some prep work was added for value checking `state`.
* feat: value checking `state` on flag update
The state should be one of the constants defined earlier in the file.
Barış Soner Uşaklı
Opliko
Julian Lam
psychobunny
Peter Jaszkowiak
Jonson Petard
hb90dev
Barış Soner Uşaklı
arite