OGM-1229 ORM 5.2 upgrade #872
Conversation
gsmet
force-pushed
the
orm-5.2-upgrade
branch
3 times, most recently
from
640dbd7
to
8d5884f
Compare
|
@gsmet the JTA failure on Neo4j should be solved with the ORM uprade right? |
|
@Sanne yes, the JTA issue is fixed. But I have another one with the WF 11 upgrade. |
|
Not urgent but for me to have a look it would be useful to rebase it first. Let me know if I can help with the WF11 upgrade? I tried the current branch but it's attempting to load a non-existing org.hibernate:hibernate-orm-modules:zip:5.2.11.Final with classifier wildfly-10-dist so I guess we're not fully in synch |
|
Yes, I upgraded to Final, then noticed the WF11 upgrade, then had to upgrade to Search 5.8 and so on :). I'm working on fixing (hopefully) last issue. Anyway, we shouldn't merge it for now, I think. Better apply it when we initialize 5.3. |
cool
Sure, I'd still like to have a peek though :) Also feel free to identify some improvements which could be extracted and merged already. |
gsmet
force-pushed
the
orm-5.2-upgrade
branch
from
a712404
to
ac17d74
Compare
|
@Sanne rebased! |
gsmet
force-pushed
the
orm-5.2-upgrade
branch
2 times, most recently
from
23d22d3
to
b6ad0f3
Compare
|
Assigning to @emmanuelbernard as it's just some maintenance, a trivial minor upgrade :D |
|
Jenkins, retest this please |
| * @return a new set of entities | ||
| */ | ||
| private static Set<Class<?>> toRootEntities(ExtendedSearchIntegrator searchFactoryImplementor, Class<?>... selection) { | ||
| private static IndexedTypeSet toRootEntities(ExtendedSearchIntegrator extendedIntegrator, Class<?>... selection) { |
There was a problem hiding this comment.
Just to let you know I noticed this, and yes this is going to be much better with the next round of the Search SPI
gsmet
force-pushed
the
orm-5.2-upgrade
branch
from
b6ad0f3
to
6b1e953
Compare
gsmet
force-pushed
the
orm-5.2-upgrade
branch
2 times, most recently
from
8cefb8c
to
9029b6f
Compare
|
Tried to align the versions with WildFly 11 but I had an issue with Narayana: https://developer.jboss.org/message/976535#976535 . Worst case scenario, we can keep an older version for the tests but let's see if we can figure it out with the help of the Narayana people. |
|
The issue with Narayana was fixed thanks to Tom Jenkinson. I rebased the branch onto latest master. |
gsmet
force-pushed
the
orm-5.2-upgrade
branch
from
9a0e552
to
52a8ca7
Compare
|
Rebased and upgraded Search to 5.8.1.Final. |
gsmet
force-pushed
the
orm-5.2-upgrade
branch
from
52a8ca7
to
ac0dd33
Compare
gsmet
force-pushed
the
orm-5.2-upgrade
branch
2 times, most recently
from
6194ce6
to
89567c9
Compare
…onImpl is now an EntityManager Also removes a lot of copied code to be able to determine what is strictly necessary. Note that this code will be reintroduced later (and then in the end removed due thanks to some changes in ORM). Note that this commit removes some APIs (NoSQLQuery and NoSQLQueryImplementor) as they are not necessary anymore and requires us to maintain more things. We discussed it with Sanne and, considering the fact that ORM 5.2 breaks a lot of things anyway, it would be considered acceptable.
This is due to the fact that Session and EntityManager are now a same thing so Session has to follow JPA requirements.
…ntityManager and it is created as an OGM session
…n class It was already the case but as Session is also an EntityManager now, we have more methods in it (and the code in ORM has more private/protected methods we need to copy). Note that the methods were deleted in a previous commit to know exactly what was required. They will be removed in a subsequent commit relying on the new CustomLoaderFactory mechanism introduced in ORM.
…n() in a JTA environment anymore A lot of our shared tests rely on this and we want to keep them working for Neo4j.
It was required for a moment in my journey and I think it's more future proof to have it there and rely on ORM mechanisms.
Note that it is not an incompatibility in ORM: the method was introduced during the 5.2.11 release cycle and the name changed before the release.
ORM used to initialize a TransactionImpl when initializing the SessionImpl but it now only calls pulse() on the TransactionCoordinator. Thus, the OperationCollector initialization must be moved there.
gsmet
force-pushed
the
orm-5.2-upgrade
branch
from
9c577d6
to
a97bb73
Compare
|
Jenkins, rebuild this please |
|
@DavideD just force pushed a fix. I didn't have the infinispan-embedded project in my workspace so I didn't see the compilation issue with the new added test. Should be fine now. |
|
Merged #872 Thanks |
|
@DavideD thanks! Glad to see this one finally in. |
Upgrade to ORM 5.2 and Search 5.8.
One thing that we might need to discuss is the removal of NoSQLQuery. It might cause some issues on the user side but relying on NativeQuery (which now has a good name - it's not called SQLQuery anymore) makes it far easier for the integration. If we do want to keep NoSQLQuery, we will need to add a QueryFactory facility to the ORM Session.