7.0.0-RC2 BUG FIX - for properties during PreUpdateEvents not being marked dirty when using mongodb

[codeconsole]

With mongodb, anything done during a PreUpdateEvent is lost due to not being marked dirty.

This fix wraps FieldEntityAccess with ModificationTrackingEntityAccess then marks the modified properties as dirty.
Hibernate does the same thing, but it is still not necessary for hibernate to do so in order for properties to be marked dirty.

# clone grails-core with the fix and pTML
git clone --depth 1 --branch 7.0.x-mongo-pre-update-dirty-fix https://github.com/codeconsole/grails-core grails-core-mongo-pre-update-dirty-fix && cd grails-core-mongo-pre-update-dirty-fix && ./gradlew pTML

# Start mongo  and hibernate versions with the fix
git clone --depth 1 --branch mongodb-mavenLocal-7.0.0-SNAPSHOT https://github.com/codeconsole/mongo-test mongo-test-mongo-SNAPSHOT && cd mongo-test-mongo-SNAPSHOT && ./gradlew bootRun --args='--server.port=8081'

git clone --depth 1 --branch h2-mavenLocal-7.0.0-SNAPSHOT https://github.com/codeconsole/mongo-test mongo-test-hibernate-SNAPSHOT && cd mongo-test-hibernate-SNAPSHOT && ./gradlew bootRun --args='--server.port=8082'

Verify fix:
http://localhost:8081/user/create
http://localhost:8082/user/create

[matrei]

I think this is a possible candidate for 7.0.1.
We need to ship 7.0.0 to lay some stable ground so we can continue moving from there.

[codeconsole]

@matrei but that is the reason behind this PR. Mongo is broken. This PR only affects mongo. Should 7.0.0 really go out broken?

[matrei]

@matrei but that is the reason behind this PR. Mongo is broken. This PR only affects mongo. Should 7.0.0 really go out broken?

Yes, I'm afraid that would be my suggestion. I don't think we can let any more time go by. I think we need to ship 7 now.
It's almost a year since I wrote this comment apache/grails-data-mapping#1855 (comment), and my gut feeling seems to have been right at that time.

[jdaugherty]

Since this is affecting mongo only now, I do not have a concern with making this change if the tests pass. I'm ok with it also going in 7.0.1 since it's limited in scope / wont' affect hibernate.

[codeconsole]

Since this is affecting mongo only now, I do not have a concern with making this change if the tests pass. I'm ok with it also going in 7.0.1 since it's limited in scope / wont' affect hibernate.

@jdaugherty This will not fix beforeUpdate() {}. This will fix EventListeners.

[codeconsole]

Yes, I'm afraid that would be my suggestion. I don't think we can let any more time go by. I think we need to ship 7 now. It's almost a year since I wrote this comment apache/grails-data-mapping#1855 (comment), and my gut feeling seems to have been right at that time.

@matrei what does your previous comment have to do with mongo being broken? There is no issue with the @AutoTimestamp code. It has worked fine with hibernate. The issue is due to a specific bug with mongo.

This PR has nothing to do with @AutoTimestamp.

[matrei]

what does your previous comment have to do with mongo being broken? There is no issue with the @AutoTimestamp code. It has worked fine with hibernate. The issue is due to a specific bug with mongo.

Sorry, my mistake. I may have misunderstood or mixed up your PRs.
Thank you for creating tests, it saves so much time understanding the bug and reviewing the bug fix.

[jdaugherty]

@codeconsole I believe this can be closed now that the other PR was merged?

[codeconsole]

no longer needed due to #15143