closes#1873.
During file system merge upgrades of new releases, old files are not removed and node's require loads the old file instead of all the new ones in the new directory. The files in this commit act as a delegate for all other dependent scripts. These shim files explicitly require the new index.js.
fixes#1351
- when a post has a published_at value show a blank
placeholder
- when a post doesn’t have a published_at
value then show the required published at value
format
closes#1880
- added `rss` to core/server/models `generateSlug` reserved keywords
- added integration test for safe slug generation to core/test/integration/model/model_posts_spec.js
The tags UI is pretty messed up all round. This fix makes it work properly at least on desktop. Mobile needs a lot more love. This is a followup to #1774 and most problems stemming originally from #710
closes#1464
- adds opt-out via updateCheck:false in config.js
- update check is done on admin index, but doesn't interfere with rendering
- adds update check module, which gets the usage data, makes the request and handles the response
- adds two new settings to default-settings, one for next check time, and one for whether to show the notification
- adds a new rejectError method to errorHandling
- adds a new helper for displaying the notification
Conflicts:
core/server/helpers/index.js
core/test/unit/server_helpers_index_spec.js
#1351
- prevent a new post (not saved on server) from
updating its slug/date to the server
- fix jshint
- add back creation of a posts slug upon a post creation
- update for rebasing
- hide ability to ‘delete this post’ from post settings
menu when a post hasn’t yet been saved to the server
closes#1854
- added blueimp file upload to debug.js
- changed POST /ghost/api/v0.1/db to be used with AJAX
- cache invalidation header should now work for import
- moved busboy middleware invocation to routes/api and routes/admin
- moved api.db.import to api.db.importContent (I hated the [] notation)
- moved api.db.export to api.db.exportContent (see above)
- use hex instead of base64 as this can cause errors when trying to reopen the file due to characters like '/' appearing
- added basic console log to errors.
issue #1378fixes#1328
- xss santization does some odd things. This isn't needed until we have multi-user support, and we are investigating better solutions.
fixes#1765fixes#1811
issue #1833
New UrlFor functions
- moved body of url helper to config.path.urlFor, which can generate a URL for various scenarios
- urlFor can take a string (name) or object (relativeUrl: '/') as the first
argument - this is the first step towards issue #1833
- also added config.path.urlForPost which is async and handles getting
permalink setting
- frontend controller, ghost_head helper, cache invalidation all now use
urlFor or urlForPost all urls should be correct and consistent
URL Consistency Improvements
- refactored invalidateCache into cacheInvalidationHeader which returns a
promise so that url can be generated properly by urlForPost
- moved isPost from models to schema, and refactored schema to have a tables object
- deleted posts now return the whole object, not just id and slug,
ensuring cache invalidation header can be set on delete
- frontend controller rss and archive page redirects work properly with subdirectory
- removes {{url}} helper from admin and client, and replaced with adminUrl
helper which also uses urlFor
- in res.locals ghostRoot becomes relativeUrl, and path is removed
closes#1801
- adding fixBools method to `server/models/base.js` to convert bools to 1/0 to be consistent with MySQL & sqlite3 data providers (based on @ErisDS recommendation)
- this in turn fixes the check in `server/controllers/frontend.js`, which does an explicit `post.page === 0` comparison (in pgsql this is a `boolean`, since the schema declares it as "bool" in `server/data/schema.js`, but MySQL/sqlite3 don't have concept of Boolean, only an integer or bit)
- any model retrieved from persistence will pass through this (possible future refactoring is to combine fixBools & fixDates into one "canonicalize()" to have a single loop pass)
throttle list
- once a user has successfully logged into ghost
they no longer are a malicious user and as such
their IP address should be removed from the
array of login attempts
- should also reduce the memory usage of Ghost
as the loginSecurity array gets pruned upon
every successful login
- this also fixes a race condition i was experiencing
during functional tests wherein i would receive
the login throttle message during regular testing.
Seems my machine is able to run casper fast enough
that it could complete each test under an amount
of time that tripped the login throttle message.
