Releases: cs3org/reva
v2.27.3
Changelog for reva 2.27.3 (2025-01-27)
The following sections list the changes in reva 2.27.3 relevant to
reva users. The changes are ordered by importance.
Summary
Details
- Bugfix #5042: Fix ocis dependency
Fix the ocm gateway connection pool. Fix ocis dependency in the reva go.mod file. Bump the ocis
version accordingly to the major version.
owncloud/ocis#10846
owncloud/ocis#10878
#5042
- Enhancement #5051: Emit SpaceMembershipExpired event
v2.27.2
Changelog for reva 2.27.2 (2025-01-17)
The following sections list the changes in reva 2.27.2 relevant to
reva users. The changes are ordered by importance.
Summary
- Fix #5043: Fix SpaceMembershipExpired unmarshal
Details
- Bugfix #5043: Fix SpaceMembershipExpired unmarshal
SpaceMembershipExpired events can now be unmarshalled correctly
v2.27.1
Changelog for reva 2.27.1 (2025-01-09)
The following sections list the changes in reva 2.27.1 relevant to
reva users. The changes are ordered by importance.
Summary
- Fix #5033: Fix ocm wildcards
- Fix #5031: Allow to accepted invite after it was once deleted
- Fix #5026: Delete Blobs when Space is deleted
- Enh #5025: Allow wildcards in OCM domains
- Enh #5023: Notification feature toggle
- Enh #4990: Allow locking via ocm
- Enh #5032: Add SendEmailsEvent
Details
- Bugfix #5033: Fix ocm wildcards
Ocm wildcards were not working properly. We now overwrite the wildcard values with the actual
domain.
- Bugfix #5031: Allow to accepted invite after it was once deleted
Allowed to accepted invite even after it was once deleted on the invite receiver or invite
creation side.
- Bugfix #5026: Delete Blobs when Space is deleted
Delete all blobs of a space when the space is deleted.
- Enhancement #5025: Allow wildcards in OCM domains
When verifiying domains, allow wildcards in the domain name. This will not work when using
verify-request-hostname
- Enhancement #5023: Notification feature toggle
Adds a feature toggle for the notification settings.
- Enhancement #4990: Allow locking via ocm
Implement locking endpoints so files can be locked and unlocked via ocm.
- Enhancement #5032: Add SendEmailsEvent
Adds SendEmailsEvent that is used to trigger the sending of group emails.
v1.29.0
Changelog for reva 1.29.0 (2025-01-07)
The following sections list the changes in reva 1.29.0 relevant to
reva users. The changes are ordered by importance.
Summary
- Fix #4898: Make ACL operations work over gRPC
- Fix #4667: Fixed permission mapping to EOS ACLs
- Fix #4520: Do not use version folders for xattrs in EOS
- Fix #4599: Auth: increase verbosity of oidc parsing errors
- Fix #5006: Blocking reva on listSharedWithMe
- Fix #4557: Fix ceph build
- Fix #5017: No empty favs attr
- Fix #4620: Fix ulimits for EOS container deployment
- Fix #5015: Fixed error reporting in the EOS gRPC client
- Fix #4931: Fixed tree metadata following fix in EOS
- Fix #4930: Make removal of favourites work
- Fix #4574: Fix notifications
- Fix #4790: Ocm: fixed domain not having a protocol scheme
- Fix #4849: Drop assumptions about user types when dealing with shares
- Fix #4894: No certs in EOS HTTP client
- Fix #4810: Simplified error handling
- Fix #4973: Handle parsing of favs over gRPC
- Fix #4901: Broken PROPFIND perms on gRPC
- Fix #4907: Public links: return error when owner could not be resolved
- Fix #4591: Eos: fixed error reporting for too large recycle bin listing
- Fix #4896: Fix nilpointer error in RollbackToVersion
- Fix #4905: PurgeDate in ListDeletedEntries was ignored
- Fix #4939: Revert 'make home layout configurable'
- Enh #5028: Handle empty EOS traces
- Enh #4911: Cephfs refactoring + make home layout configurable
- Enh #4937: @labkode steps down as project owner
- Enh #4579: Remove domain-specific code to other repos
- Enh #4824: Refactor Ceph code
- Enh #4797: Refactor CI jobs and bump to latest deps
- Enh #4934: Access to EOS via tokens over gRPC
- Enh #4870: Only load X509 on https
- Enh #5014: Log app when creating EOS gRPC requests
- Enh #4892: Do not read eos user ACLs any longer
- Enh #4720: Differentiate quota for user types in EOS
- Enh #4863: Favourites for eos/grpc
- Enh #5013: Updated dependencies + moved to go 1.22
- Enh #4514: Pass lock holder metadata on uploads
- Enh #4970: Improved logging on createHome
- Enh #4984: Drop shadow namespaces
- Enh #4670: Ocm: support bearer token access
- Enh #4977: Do not use root on EOS
Details
-
Bugfix #4898: Make ACL operations work over gRPC
This change solves two issues: * AddACL would fail, because the current implementation of
AddACL in the EOS gRPC client always sets msg.Recursive = true. This causes issues on the EOS
side, because it will try running a recursive find on a file, which fails. * RemoveACL would
fail, because it tried matching ACL rules with a uid to ACL rules with a username. This PR changes
this approach to use an approach similar to what is used in the binary client: just set the rule
that you want to have deleted with no permissions. -
Bugfix #4667: Fixed permission mapping to EOS ACLs
This is to remove "m" and "q" flags in EOS ACLs for regular write shares (no re-sharing).
-
Bugfix #4520: Do not use version folders for xattrs in EOS
This was a workaround needed some time ago. We revert now to the standard behavior, xattrs are
stored on the files. -
Bugfix #4599: Auth: increase verbosity of oidc parsing errors
This is to help further debugging of auth issues. An unrelated error reporting was also fixed.
-
Bugfix #5006: Blocking reva on listSharedWithMe
listSharesWithMe
blocked a reva thread in the case that one of the shares was not resolvable.
This has now been fixed -
Bugfix #4557: Fix ceph build
-
Bugfix #5017: No empty favs attr
See issue #5016: we now unset the favs attr if no more favs are set
-
Bugfix #4620: Fix ulimits for EOS container deployment
-
Bugfix #5015: Fixed error reporting in the EOS gRPC client
This in particular fixes the lock-related errors
-
Bugfix #4931: Fixed tree metadata following fix in EOS
The treecount is now populated from the EOS response.
-
Bugfix #4930: Make removal of favourites work
Currently, removing a folder from your favourites is broken, because the handleFavAttr
method is only called in SetAttr, not in UnsetAttr. This change fixes this. -
Bugfix #4574: Fix notifications
-
Bugfix #4790: Ocm: fixed domain not having a protocol scheme
This PR fixes a bug in the OCM open driver that causes it to be unable to probe OCM services at the
remote server due to the domain having an unsupported protocol scheme. in this case domain
doesn't have a scheme and the changes in this PR add a scheme to the domain before doing the probe. -
Bugfix #4849: Drop assumptions about user types when dealing with shares
We may have external accounts with regular usernames (and with null uid), therefore the
current logic to heuristically infer the user type from a grantee's username is broken. This PR
removes those heuristics and requires the upper level to resolve the user type. -
Bugfix #4894: No certs in EOS HTTP client
Omit HTTPS cert in EOS HTTP Client, as this causes authentication issues on EOS < 5.2.28. When
EOS receives a certificate, it will look for this cert in the gridmap file. If it is not found
there, the whole authn flow is aborted and the user is mapped to nobody. -
Bugfix #4810: Simplified error handling
Minor rewording and simplification, following cs3org/OCM-API#90 and cs3org/OCM-API#91
-
Bugfix #4973: Handle parsing of favs over gRPC
To store user favorites, the key
user.http://owncloud.org/ns/favorite
maps to a list of
users, in the formatu:username=1
. Right now, extracting the "correct" user doesn't happen
in gRPC, while it is implemented in the EOS binary client. This feature has now been moved to the
higher-level call in eosfs. -
Bugfix #4901: Broken PROPFIND perms on gRPC
When using the EOS gRPC stack, the permissions returned by PROPFIND on a folder in a project were
erroneous because ACL permissions were being ignored. This stems from a bug in
grpcMDResponseToFileInfo, where the SysACL attribute of the FileInfo struct was not being
populated.#4901
see: -
Bugfix #4907: Public links: return error when owner could not be resolved
-
Bugfix #4591: Eos: fixed error reporting for too large recycle bin listing
EOS returns E2BIG, which internally gets converted to PermissionDenied and has to be properly
handled in this case. -
Bugfix #4896: Fix nilpointer error in RollbackToVersion
-
Bugfix #4905: PurgeDate in ListDeletedEntries was ignored
The date range that can be passed to ListDeletedEntries was not taken into account due to a bug in
reva: the Purgedate argument was set, which only works for PURGE requests, and not for LIST
requests. Instead, the Listflag argument must be used. Additionally, there was a bug in the
loop that is used to iterate over all days in the date range. -
Bugfix #4939: Revert 'make home layout configurable'
Partial revert of #4911, to be re-added after more testing and configuration validation. The
eoshome vs eos storage drivers are to be adapted. -
Enhancement #5028: Handle empty EOS traces
-
Enhancement #4911: Cephfs refactoring + make home layout configurable
-
Enhancement #4937: @labkode steps down as project owner
Hugo (@labkode) steps down as project owner of Reva.
-
Enhancement #4579: Remove domain-specific code to other repos
-
Enhancement #4824: Refactor Ceph code
-
Enhancement #4797: Refactor CI jobs and bump to latest deps
-
Enhancement #4934: Access to EOS via tokens over gRPC
As a guest account, accessing a file shared with you relies on a token that is generated on behalf
of the resource owner. This method, GenerateToken, has now been implemented in the EOS gRPC
client. Additionally, the HTTP client now takes tokens into account. -
Enhancement #4870: Only load X509 on https
Currently, the EOS HTTP Client always tries to read an X509 key pair from the file system (by
default, from /etc/grid-security/host{key,cert}.pem). This makes it harder to write unit
tests, as these fail when this key pair is not on the file system (which is the case for the test
pipeline as well).This PR introduces a fix for this problem, by only loading the X509 key pair if the scheme of the
EOS endpoint is https. Unit tests can then create a mock HTTP endpoint, which will not trigger
the loading of the key pair. -
Enhancement #5014: Log app when creating EOS gRPC requests
-
Enhancement #4892: Do not read eos user ACLs any...
v2.27.0
Changelog for reva 2.27.0 (2024-12-12)
The following sections list the changes in reva 2.27.0 relevant to
reva users. The changes are ordered by importance.
Summary
- Fix #4985: Drop unneeded session locks
- Fix #5000: Fix ceph build
- Fix #4989: Deleting OCM share also updates storageprovider
- Enh #4998: Emit event when an ocm share is received
- Enh #4996: Get rid of some cases of unstructured logging
Details
- Bugfix #4985: Drop unneeded session locks
We no longer lock session metadada files, as they are already written atomically.
- Bugfix #5000: Fix ceph build
- Bugfix #4989: Deleting OCM share also updates storageprovider
When remvoving an OCM share we're now also removing the related grant from the storage
provider.
- Enhancement #4998: Emit event when an ocm share is received
- Enhancement #4996: Get rid of some cases of unstructured logging
v2.26.8
Changelog for reva 2.26.8 (2024-12-04)
The following sections list the changes in reva 2.26.8 relevant to
reva users. The changes are ordered by importance.
Summary
- Fix #4983: Delete stale shares in the jsoncs3 share manager
- Fix #4963: Fix name and displayName in an ocm
- Fix #4968: Jsoncs3 cache fixes
- Fix #4928: Propagate lock in PROPPATCH
- Fix #4971: Use manager to list shares
- Enh #4978: We added more trace spans in decomposedfs
- Enh #4921: Polish propagation related code
Details
- Bugfix #4983: Delete stale shares in the jsoncs3 share manager
The jsoncs3 share manager now properly deletes all references to removed shares and shares
that belong to a space that was deleted
- Bugfix #4963: Fix name and displayName in an ocm
Fixed name and displayName in an ocm PROPFIND response
- Bugfix #4968: Jsoncs3 cache fixes
The jsoncs3 share manager now retries persisting if the file already existed and picks up the
etag of the upload response in all cases.
- Bugfix #4928: Propagate lock in PROPPATCH
Clients using locking (ie. Windows) could not create/copy files over webdav as file seemed to
be locked.
- Bugfix #4971: Use manager to list shares
When updating a received share the usershareprovider now uses the share manager directly to
list received shares instead of going through the gateway again.
- Enhancement #4978: We added more trace spans in decomposedfs
- Enhancement #4921: Polish propagation related code
We polished some corner cases for propagation that reduce log messages in normal operation.
v2.26.7
Changelog for reva 2.26.7 (2024-11-20)
The following sections list the changes in reva 2.26.7 relevant to
reva users. The changes are ordered by importance.
Summary
- Fix #4964: Fix a wrong error code when approvider creates a new file
Details
- Bugfix #4964: Fix a wrong error code when approvider creates a new file
We fixed a problem where the approvider would return a 500 error instead of 403 when trying to
create a new file in a read-only share.
v2.26.6
Changelog for reva 2.26.6 (2024-11-19)
The following sections list the changes in reva 2.26.6 relevant to
reva users. The changes are ordered by importance.
Summary
- Fix #4955: Allow small clock skew in reva token validation
- Fix #4929: Fix flaky posixfs integration tests
- Fix #4953: Avoid gateway panics
- Fix #4959: Fix missing file touched event
- Fix #4933: Fix federated sharing when using an external identity provider
- Fix #4935: Enable datatx log
- Fix #4936: Do not delete mlock files
- Fix #4954: Prevent a panic when logging an error
- Fix #4956: Improve posixfs error handling and logging
- Fix #4951: Pass the initialized logger down the stack
Details
- Bugfix #4955: Allow small clock skew in reva token validation
Allow for a small clock skew (3 seconds by default) when validating reva tokens as the different
services might be running on different machines.
- Bugfix #4929: Fix flaky posixfs integration tests
We fixed a problem with the posixfs integration tests where the in-memory id cache sometimes
hadn't caught up with the cleanup between test runs leading to flaky failures.
- Bugfix #4953: Avoid gateway panics
The gateway would panic if there is a missing user in the context. Now it errors instead.
- Bugfix #4959: Fix missing file touched event
We have fixed an issue where the file touched
event was not being triggered when an office
document was created.
- Bugfix #4933: Fix federated sharing when using an external identity provider
We fixes and issue that caused federated sharing to fail when the identity provider url did not
match the federation provider url.
- Bugfix #4935: Enable datatx log
We now pass a properly initialized logger to the datatx implementations, allowing the tus
handler to log with the same level as the rest of reva.
- Bugfix #4936: Do not delete mlock files
To prevent stale NFS file handles we no longer delete empty mlock files after updating the
metadata.
- Bugfix #4954: Prevent a panic when logging an error
We fixed a panic when constructing a path failed to get the parent for a node.
- Bugfix #4956: Improve posixfs error handling and logging
We improved error handling and logging in the posixfs storage driver.
- Bugfix #4951: Pass the initialized logger down the stack
We now make the initialized logger available to grpc services and storage drivers, which
allows for easier and more uniform logging.
v2.26.5
Changelog for reva 2.26.5 (2024-11-12)
The following sections list the changes in reva 2.26.5 relevant to
reva users. The changes are ordered by importance.
Summary
- Fix #4926: Make etag always match content on downloads
- Fix #4920: Return correct status codes for simple uploads
- Fix #4924: Fix sync propagation
- Fix #4916: Improve posixfs stability and performance
Details
- Bugfix #4926: Make etag always match content on downloads
We added an openReaderfunc to the Download interface to give drivers a way to guarantee that the
reader matches the etag returned in a previous GetMD call.
- Bugfix #4920: Return correct status codes for simple uploads
Decomposedfs now returns the correct precondition failed status code when the etag does not
match. This allows the jsoncs3 share managers optimistic locking to handle concurrent writes
correctly
- Bugfix #4924: Fix sync propagation
Fixes the defers in the sync propagation.
- Bugfix #4916: Improve posixfs stability and performance
The posixfs storage driver saw a number of bugfixes and optimizations.
v2.26.4
Changelog for reva 2.26.4 (2024-11-07)
The following sections list the changes in reva 2.26.4 relevant to
reva users. The changes are ordered by importance.
Summary
Details
- Bugfix #4917: Fix 0-byte file uploads
We fixed an issue where 0-byte files upload did not return the Location header.
- Bugfix #4918: Fix app templates
We fixed the app templates by using the product name of the providerinfo instead of the provider
name.