The diff can be checked using: git diff --ignore-all-space --ignore-blank-lines
To see only the changes not related to blank lines.
Signed-off-by: Côme Chilliet <come.chilliet@nextcloud.com>
Some (well all except sqlite) database platforms support timezone
configuration. The problem is that we expect everything in UTC,
but some servers might have set some different default (e.g. in database
configuration or even just because of `TZ` environment variable).
This causes incorrect values when expecting `NOW()` to return the
current time in UTC.
For PHP we already enforce UTC as timezone, this PR adds a middleware
that enforces UTC also as the database connection / session timezone.
Signed-off-by: Ferdinand Thiessen <opensource@fthiessen.de>
Fix possible dead locks when running the propagator caused by two
requests updating the same amount rows in transactions.
- Lock rows always in the same deterministic order by sorting the
path_hash first
- On all database outside of sqlite, also do first a SELECT FOR UPDATE
to lock all the rows used in batch UPDATE calls, afterward to decrease
the risk of two requests trying to lock the same rows
Signed-off-by: Carl Schwan <carlschwan@kde.org>
We do not support Oracle 11 anymore but at least Oracle 12c (12.2).
So the limitation is gone (Oracle now supports up to 128 character long
names).
Instead we are now limited by MySQL (64 characters) and PostgreSQL (63
characters).
Signed-off-by: Ferdinand Thiessen <opensource@fthiessen.de>
- fetch is replaced with fetchAssociative/fetchNumeric/fetchOne with
better type hinting
- Same with fetchAll
- And add iterateAssociative/iterateNumeric which are nicer to use than
a `while ($row = $result->fetch()) {}`
Signed-off-by: Carl Schwan <carl.schwan@nextcloud.com>
Split the DB requests in chunk of 1000 and use INNER JOIN instead of IN
as this is better supported on all DB.
Signed-off-by: Carl Schwan <carl.schwan@nextcloud.com>
We disabled them because they are not supported on Oracle DB and it is
still the case for OCI < 23. But instead of disabling the support
completely for every database types, mark non-nullable boolean column as
actually nullable when using Oracle.
This allow to use some slighly lighter schema on normal databases who
support natively booleans wheen we don't need to store 3 states
true|false|null.
Signed-off-by: Carl Schwan <carl.schwan@nextcloud.com>
There are some behavioral differences that apps may need to check for.
See discussion on #51175 for more info.
This preserves the existing behavior of getDatabaseProvider()
Signed-off-by: Varun Patil <varunpatil@ucla.edu>
During a reconnect we are losing the connection and when the
realLastInsertId call is the one triggering the reconnect, it
does not return the ID. But inside the reconnect, we were able
to save the last insert id, so calling it a second time is going
to be successful.
We can not return the result on the initial call, as we are already
way deeper in the stack performing the actual database query on
the doctrine driver.
Signed-off-by: Joas Schilling <coding@schilljs.com>
Git'Fellow
Carl Schwan
Côme Chilliet
Benjamin Gaussorgues
provokateurin
Ferdinand Thiessen
grnd-alt
Andy Scherzinger
Carl Schwan
Joas Schilling
Tobias Kaminsky
Carl Schwan
Varun Patil
Joas Schilling
Robin Appelman
Marcel Klehr
Marcel Müller