* add new configuration setting GROUPWARE_SEND_DURATIONS_RESPONSE
(defaults to false)
* keep track of lists of durations of backend calls
* when enabled, report them as response headers Durations (human
readable) and Durations-Nanos (as raw nanosecond values for machine
consumption)
* refactor APIs in JMAP and Groupware in order to implement pagination
across multiple accountIds and multiple suppliers (currently
implemented using a mock supplier for contacts)
* requires go 1.26 due to use of self-reflecting generics type
constraints
* still missing: query criteria and sorting parameters
* still missing: multi-accountId support for emails
* errors are now all just 'error' in the APIs, instead of the
specialized implementations, and are interpreted dynamically where
necessary in order to transform them into HTTP responses
* remove position, anchor, anchorOffset as individual query parameters
as we now only support a 'next=...' token for subsequent pages
(except in emails for now), and use jmap.QueryParams instead; those
tokens have a header character for the format, followed by a JSON
encoded QueryParams map, all wrapped into base62 to make it clearer
that it is meant to be an opaque token, and not a parameter clients
should tinker with or construct themselves
* introduce QueryParamsSupplier as an interface to provide QueryParams
for various scenarios (single supplier, multiple supplier, ...) per
accountId
* implement multi-supplier template methods slist and squery
* in the JMAP API as well as in several places in the Groupware
framework, use a single jmap.Response[T] object to return the
payload, the language, the session state and the etag/state instead
of individual multi-valued return values
* the "position" attribute is always set to 0 when using an anchor for
pagination, but it does not accurately reflect the position and thus
we decided to remove the attribute from the resulting object for
clarity
* omit "canCalculateChanges" attribute when its value is "true", which
is going to be the case in 100% of calls against Stalwart
* JMAP query limit of 0 is synonymous with "no limit", but we actually
want to be able to perform queries without any results, for cases
where we only want to count the total number of objects, and also
because it makes more sense semantically
* introduce query parameter validation checks, in order to only allow
query parameters that are actually supported, which is going to be
useful during development of clients
* adds creating addressbooks, calendars, mailboxes
* adds deleting mailbox, event, identity
* adds modifying an email
* introduce template functions for the Groupware API in templates.go,
and use those in route function implementations whenever possible
* add capability checking for mail, quota, blobs
* adds Changes interface
* adds JmapResponse interface
* make the JMAP internal API a bit more future-proof by passing
ObjectType objects instead of the JMAP namespaces
* remove the new attempt to contain operations even further using the
Factory objects
* move CalendarEvent operations to its own file, like everything else
* fix email tests
* ignore WS error when closing an already closed connection
* move ContactCard from jscontact to jmap, as it is actually a JMAP
specification item, but also was causing too many issues with
circular references from jscontact -> jmap
* introduce Foo, Idable, GetRequest, GetResponse, etc... types and
generics
* first attempt at a Foo factory type for Mailboxes, needs to be
expanded to further minimize repetition
* add more specialized template functions to avoid repetition
* introduce ChangesTemplate[T] for *Changes structs
* add Groupware APIs for creating and deleting addressbooks
* add Groupware APIs for creating and deleting calendars
* add JMAP APIs for creating and deleting addressbooks, calendars
* add JMAP APIs to retrieve Principals
* fix API tagging
* move addressbook JMAP APIs into its own file
* move addressbook Groupware APIs into its own file
* implement ContactCard retrieval endpoint for syncing
* re-implement that endpoint for Email too
* fix the Mailbox changes endpoint to actually return changes about
Mailboxes, and not about Emails
* when querying the diff of Mailboxes without any prior state, return
an error since the result is not what one would expect
* introduce the 'changes' API tag and group
* refactor the successful response functions to consistently return an
object type and object state whenever possible
* move the syncing endpoints under /accounts/*/changes/ for better
clarity, e.g. /changes/emails instead of /emails/mailbox/*/changes
- fix a bunch of minor issues and typos that were found using GoLand
and gosec
- add a gosec Makefile target for Groupware related files, in
services/groupware/Makefile
- enable checking JMAP session capabilities for events and contacts,
and only enable skipping that check for tasks until those are
implemented in Stalwart as well
- fix a CWE-190 (integer overflow or wraparound) found by gosec
- consistently use struct references for methods of Groupware and
Request, instead of mixing up references and copies
- always log errors when unable to register a Prometheus metric
* re-implement the auth-api service to authenticate Reva tokens
following the OIDC Userinfo endpoint specification
* pass the context where necessary and add an authenticator interface
to the JMAP HTTP driver, in order to select between master
authentication (which is used when GROUPWARE_JMAP_MASTER_USERNAME and
GROUPWARE_JMAP_MASTER_PASSWORD are both set) and OIDC token
forwarding through bearer auth
* add Stalwart directory configuration "idmoidc" which uses the
OpenCloud auth-api service API (/auth/) to validate the token it
received as bearer auth from the Groupware backend's JMAP client,
using it as an OIDC Userinfo endpoint
* implement optional additional shared secret to secure the Userinfo
service, as an additional path parameter
* add example generator infrastructure, with some examples for pkg/jmap
and pkg/groupware, with more needing to be done
* alter the apidoc Makefile to stop using go-swagger but, instead, use
the openapi.yml file that must be dropped into that directory using
groupware-apidocs (will improve the integration there later)
* add Makefile target to generate examples
* bump redocly from 2.4.0 to 2.14.5
* introduce Request.PathParam() and .PathParamDoc() to improve API
documentation, as well as future-proofing
* improve X-Request-ID and Trace-Id header handling in the middleware
by logging it safely when an error occurs in the middleware
* fix a bug in how email summaries are flattened across multiple
accounts, which was previous resulting in empty email objects
* allow negative offset in email pagination
* make all /emails endpoints return emails without bodies
* implement Request.AllAccountIds() to generalize the fetching (and
uniqifying) of all account IDs, which will allow us to implement
things such as "subscribed" accounts, or limiting the number of
accounts in one request
* add Account-Id response header
* add Object-Type response header
* upgrade Stalwart image for devtools/full to 0.14.1
* re-assert which features are implemented or not in 0.14.1
* refactor the integration tests yet again to make it clearer and
easier to see those "features-or-not"
* get rid of old tests that are now better covered by integration tests
* rewrite how we compare expected and actual objects in integration
tests, finally having found a way to ignore the @type attribute
properly instead of having to mutate all objects to remove it
