Last Update: 2021-05-07
Iteration of checked event series begins too late.This has been solved by considering duration when initialize recurrence iterator for conflicting series events in checked period.
Missing log message for failed authentication attempts against primary mail/transport server.This has been solved by adding logging failed authentication attempts against primary mail/transport server.
From address determined by examining user’s primary mail account data.Solution: Orderly pre-select user’s default send address when composing new mails.
Missing convert of
‘br‘ when html is preferred mode for mails.This has been solved by adding missing convert of
Button was drawn but not visible.This has been fixed by triggering ‘complete’ to adjust height again.
Wrong data from external calendar source taken over as-is.This has been fixed by adjusting bogus all-day dates prior to storing event data from subscriptions.
Too low settings for HTTP connection pools for both - auto-config server and ISPDB end-point.This has been solved by increasing settings for HTTP connection pool of both - auto-config server and ISPDB end-point - while lowering values for read and connect timeout.
SMTP host information advertised as “None” in case SMTP authentication is disabled through configuration.Now do not advertise SMTP host information as “None” in case SMTP authentication is disabled through configuration.
java.lang.StringIndexOutOfBoundsException when trying to decode subject string obtained from ENVELOPE fetch item.Fixed possible
java.lang.StringIndexOutOfBoundsException when trying to decode subject string obtained from ENVELOPE fetch item.
No fallback access used when collecting pending alarm triggers from disabled accounts.This has been solved by using fallback access when collecting pending alarm triggers from disabled accounts.
Lock entry not cleansed from database in case temporary database outage/inaccessibility occurs.This has been solved by enhancing acquired lock by a time stamp that gets periodically touched (every minute). Consider lock as expired if not touch for more than 5 minutes.
“[ ]” were not part of suffix characters we use in our regex to detect the end of links, similar to “, . ?” etc.This has been fixed by adding “[ ]” to possible suffix characters.
Capabilities were not used correctly, selectors were no longer valid and tour accidentally opened the chat app.This has been solved by adjusting selectors and capabilities and no longer open the chat app.
The local time zone was used to render the timestamp in the filter rule.This has been solved by now using UTC for rendering.
Wrong selection of day with certain (work)week settings.This has been fixed by removing basic setting dependent .startOf(‘week’) and replace with startOf(‘isoWeek’) in addition to a small adjustment for choosing the correct day.
Settings were not updated and may contain old account name.This has been solved by updating settings correctly.
A broken image link leads to failure of send/transport attempt.Solution: Don’t let failed image URI resolution prevent from sending a mail.
Existent data export tasks silently deleted if associated user/context do no more exist.This has been fixed by not deleting such “orphaned” data export tasks when invoking
listdataexports command-line tool.
A translation from a previous bugfix was missing.This has been solved by adding the missing translation.
A combination of changes and streamlining caused loss of firstname and lastname, company was used as fallback.This has been solved by adding available data were possible so company name is not used as fallback, prefer display name over company name for external participants.
Was caused by possible long-running Matcher.find() invocation.This has been fixed by adding fast plausibility check & introduced a timeout-aware matcher alternative that respects a passed timeout whenever matching the input sequence or finding a certain sub-sequence is requested to avoid possibly long-running matcher invocations.
Context names are checked case-insensitive for equality when attempting to change a context’s name and thus changing to the same context name, but different cases were considered as a no-op.This has been solved by checking case-sensitive for equal context names when attempting to change a context’s name.
No response status distinction for read-only operations in If-None-Match/If-Match checks.This has been solved by using HTTP 304 response during If-Match/If-None-Match checks for GET and HEAD.
Missing special handling for error codes that advertise actual transport succeeded, but append to standard sent folder failed.This has been fixed by adding special handling for error codes that advertise actual transport succeeded, but append to standard sent folder failed.
Inconsistent translation of ‘notes’.This has been fixed by adjusting translations.
Too many occurrences of low-level HTTP end-point pools for initialized Sproxyd clients.This has been fixed by adding cache for low-level Sproxyd HTTP end-point pools.
The filter cannot evaluate type ‘auto’ for text colors in shapes (Presentation and Spreadsheet, ooxml).Solution: Instead of sending ‘auto’ when the user selects ‘Auto’ as a text color, the best text color is evaluated corresponding to the shape background. This calculated color is sent to the filter.
When an image is inserted via the buttons in template drawings, the mousedown happens on the content root node, but the mouseup does not. But these events are registered for an optional scrolling. Therefore the scroll position was not correctly adapted when the user changes the slide using the slide pane and does not click at least once into the document after inserting the image.This has been solved by checking the target nodes for mouse down and mouse up events that are required for scrolling.
Document was not flushed before the copy was created in Drive. Flushing causes to save all pending changes which, in Spreadsheet, includes to commit the cell edit mode.Solution: Flush document before starting to copy the file in Drive for user actions “Save As” and “Save As Template”.
Templates contained more than 5 different languages on XML level.This has been fixed on XML level, replaced all (western) lang attrs to be only en-US for EN templates, de-DE for DE templates.
Event listerners were still listening on an old collection.This has been fixed by adjusting event listeners after folder rename.
Single and double clicks on the same element were competing and leaded to inconsistent behavior.This has been fixed by treating double clicks as single clicks on list elements in list layouts.
Buttons were not enabled after dialog gets idle.This has been solved by setting dialog to idle also when cropped image can’t be loaded.
Was caused by forwardUnquoted not recognized by plaintext editor.This has been solved by adding forwardUnquoted detection for plaintext editor.
Draft was saved before all delete requests were processed.This has been solved by waiting for all delete requests to be resolved, also if draft gets deleted.
Different generation of fallback attachment filename extension.This has been solved by using the common method to yield fall-back name with a reasonable file extension.
Unexpected trigger of update task for a schema that is currently checked for possible expired composition spaces.Skip clean-up of expired composition spaces for those schemas that are currently updated or need an update to solve this.
Sender’s full name for introduction in drive mail notifications escaped twice.This has been fixed by escaping sender’s full name for introduction in drive mail notifications only once.
The callback URL was not constructed properly.This has been fixed by constructing the callback URL properly.
Restricted scopes session parameter were incompatible with portable serialization.This has been solved by using comma-separated string for restricted scopes session parameter.
Equal exceptions chained multiple times.This has been solved by avoiding chaining equal exceptions multiple times.
Possible concurrent modification of storage objects is quitted with “HTTP/1.1 423 Locked” status response leading to abortion of request processing.This has been fixed by introducing retry mechanism with exponential back-off in case Sproxyd service quits request with “HTTP/1.1 423 Locked”.
Intermediate clean-up task unexpectedly dropped file storage resources.This has been solved by not running clean-up task when there are currently running data export tasks.
Primary address was used in all cases.This has been fixed by adding recipient parameter when calling api.ack and in case mail was send to alias address also this one is used for ack.
The short drag started the drag and drop mechanism of the appointment. Usually, this will not be a problem, but if the appointment is not within the displayed timeframe, the drag and drop mechanism does not work.This has been solved by disabling drag and drop for appointments, that are not within the visible timeframe.
Documentation simply was not added during development.Add documentation from feature description to technical docs.
With WebSockets disabled, desktop notifications for mail didn’t fetched a contact image.This has been solved by refactoring mail desktop notifications to use the same message style as with WebSockets enabled.
Heading ‘Reminder’ was still visible on the settings pane when showPastReminders was set to protected.Do not render heading when setting is protected.
Theme has been translated with Design and Design has also been translated with Design.This has been solved by changing the translation of Theme from Design to Theme.
Was caused by missing listener to detect, whether the vcard is attached or not.This has been fixed by introducing missing listeners.
This was caused by a missing trigger and listener for reset events.This has been solved by adding missing trigger and listener for reset events.
Settings pane for account was not updated when recovering passwords.This has been fixed by adding listener to refresh and update the account settings pane.
Due to incorrectly linked events, the portal widget does not always detect when events are changed.The incorrectly linked events were adjusted accordingly to solve this issue.
The header and footer were absolute positioned, which doesn’t look nice with a flex layout.This has been fixed by refactoring markup to use flex layout as it is intended.
ForwardUnquoted was not recognized by plaintext editor.This has been solved by adding forwardUnquoted detection for plaintext editor.
No conversion to default time zone when printing in month and week view.This has been fixed by adding time zone conversion.
UI was not updated after recovering accounts.This has been fixed by triggering an UI update, when accounts are recovered.
Due to an unhandled error the login could be prevented.Now the error is intercepted so that the login is not aborted.
Wrong timezone was selected when parsing date input.This has been solved by using default timezone when parsing the date input.
When spam folder is empty and you move a mail to it (via “mark as spam”) folder count was still 0.This has been solved by adding a refresh of the folder.
Feed sometimes wrongly uses numeric character reference instead of char.This has been solved by adding a rule to replace those occurrences with simple quotes.
Was caused by custom print rule of individual mail applies.This has been fixed by overwriting css page property.
The dropdown overlaped the viewport.Now, when overlap is detected make dropdown scrollable.
In some cases not the “total” value of a folder was used for display but a calculation. If the setting “com.openexchange.showAdmin” is set to false the displayed value differs from the actual number.If the folder supports the “total” value this value will be used now. If the setting “com.openexchange.showAdmin” is set to false, the displayed value is calculated accordingly.
Was caused by post-processing after calendar import is triggered per event group.This has been solved by importing post-process results in single task, enqueue long running import jobs.
CSS rule for overflow was overruled.This has been solved by improving selector so overflow rule is active again.
tinymce adjusts height of node flexible and “starts” with a single line.Add a dynamic min-height as it is already set for iframe container.
Firefox has od focus behavior, scrolls to bottom on focus, and ignores scroll top function.This has been solved by deferring scroll top to fix firefox focus bug.
The signature content was not correctly recognized when switching from plain text to html editor.This has been solved by removing signature on editor toggle and append it again afterward.
macOS client sends unconditional DELETE for no longer listed vCard resources after list of synchronized folders changes.Use variable path to special aggregated collection with different modes for macOS clients and introduced new modes for folders in aggregated collection.
A modified “subscribed” or “usedforsync” status in one of the underlying folders is not recognized during the incremental synchronization of the aggregated collection in CardDAV.This has been fixed by including folder state in sync-token of aggregated collection for CardDAV.
ETag and Schedule-Tag header values not submitted as quoted string.Submit ETag and Schedule-Tag header values as quoted string.
Decoded strings from basic authentication header were trimmed.Don’t trim decoded strings from basic authentication header.
Unknown timezone in invitation not interpreted correctly.More sophisticated comparison of parsed timezone observances during import.
Inefficient max. size restriction of in-memory folder cache.This has been solved by using the SessionD events when the short term sessions are removed and use the Guava cache’s expireAfterAccess method with a decent max time that should only remove stale entries.
One optimisation was done: Resume reading an S3 object’s content when HTTP connection gets unexpectedly closed due to premature EOF (actually read bytes do not match advertised content length)
Several fallback machnisms led to duplicate entries.This was fixed by avoiding to import an already existing email.
Deletion of data export task fails due to missing/absent user/context entities when querying appropriate schema reference for a user to operate on correct database.This has been solved by making config-cascade robust towards missing/absent user/context entity.
The company and department fields were not checked if they were set in the actual contact object.This has been solved by checking if the company and department are set in the actual contact before adding them to the vcard file.
Decoding with URLDecoder caused the plus sign to be converted into a space character.This has been solved by fixing the URI decoding.
Null check for relay state was not sufficient.This has been solved by properly checking for empty relay state.
In case com.openexchange.showAdmin was set to false the check for contact count was wrong.
Primary address was unnecessary checked.Don’t require primary address when checking mail account connectivity to solve this issue.
Likely a database error happens when trying to create or modify an appointment, but unfortunately the clean-up code itself raises an error that overlays the original one. Thus it is not possible to see the database error causing the failing create/update.Don’t overlay possible exceptions when performing clean-up stuff. The associated change cannot be considered as a fix for this issue. However, it is necessary to detect what is really going wrong when attempting to create or modify an appointment.
Address to notify not checked if covered by user’s aliases. if so, not notification should be sent.Do not advertise “disp_notification_to” field in a mail’s JSON representation if address to notify is covered by user’s aliases to solve this issue.
Case-sensitive check if provided E-Mail addresses are contained in set of user aliases.This has been solved by ignore-case checking if provided E-Mail addresses are contained in set of user aliases.
Was caused by wrong package name.This has been solved by using correct package name.
Standard display message advertised to client in case error “PSW-0001” (“Cannot change password…”) occurs when user’s attempts to change his/her password.This has been solved by adding better understandable display message when error code “PSW-0001” (“Cannot change password…”) is advertised by Open-Xchange Middleware.
This is caused by hundreds of reload configuration calls with each one triggering an appsuite history check.Fixed by never perform history checks in parallel.
Some images doesn’t contain a timezone in addition to the capture date. In those cases the library which extracts the capture date uses the GMT timezone as a fallback in case the timezone information is missing in the exif data.This has been solved by using the user’s timezone as a fallback for the capture date instead. Please be aware that this is still not a perfect solution for this problem. For example it depends on the timezone configuration of the appsuite when the image has been uploaded. So for example in case the timezone between the camera and the appsuite is different this leads to similar problems. Or in case the timezone of the appsuite is changed then images uploaded before and after the change have a different offset. Also this fix only applies to newly uploaded files. Existing files are still going to show the capture date based on the previous calculation which used the GMT timezone.
Excessively long-running operation to look-up a subsequence/pattern in HTML content.This has been solved by adding conditions for early abort and ultimately shield from too excessive matcher execution.
Slightly different attachment check for get and all requests. In case the content-disposition header is missing the get request in contrast to the all request considers the name attribute of the content-type header to identify attachments.This has been solved by considering the name attribute during all requests as well.
Update task accidentally removed when updating update task framework, although it was used as dependency for other tasks.This has been fixed by restoring removed update task.
“pref” parameter is used by the server to differentiate between multiple numbers of the same type, while the client only recognized one “pref”, as general preference.This has been fixed by only adding “pref” parameter when exporting TEL properties if required. Note that this is only a mitigation, e.g. when there are multiple “cell” or “home” numbers, the “pref” parameter will still be set.
Back when the sorting order was changed to descending one case was not adjusted.This has been fixed by using desc sorting order when not using imap search.
The client creates contacts in folder 6 (which is not allowed), so the server stores it in the user’s default contact folder implicitly. After a deletion of this contact in the web interface, this is only indicated for this folder, so that the client assumes that the contact in folder 6 is still there.Re-route newly created contacts to default and fake deletion in targeted collection.
Combining text nodes after pasting, so that the email-detection process finds the pasted email-addresse
In the specified environment, the comment ID was transferred to OX Presentation, but it was not found in the parameters of the application launcher. ‘Go to comment’ in notification mails is working now in SingleTab environment
z-index of the topbar was set to 2 because IE 11.
OX Documents monitors the life-cylce and online state of all Appsuite OX Documents nodes. Handling of lost Hazelcast nodes works as expected, but there a some more situations where we see the described behavior. The OX Documents monitor implementation now checks the lifecycle events from Hazelcast more carefully and detects that a merge has been done. This is handled and internal classes are re-initialized to work with the new Hazelcast uuid (especially the JMS queue names are derived from it).
Shorten title in German, check other languages if they are affected, too.
When we receive a jms message we check if all header keys are valid. If it is not the case we will stop processing the message. We change the behaviour in case there is an invalid header in the jms message. We lock this event, but we do not stop processing of this message.
A maximum runtime needs to be introduced for each call to the PDFTool (similar to watchdog for RE processes), returning an error after the configured jobExecution timeout time and responding to the appropriate request in time.
Fixed the enable/disable state of the ‘create folder’ button. Therefore, the button is not clickable in wrong cases, the error can’t happen anymore.
Be more verbose on errors, kill the pid-file.
When we must do a close hard for a document we send 3 jms messages. If one message got lost we would have a problem.